Journal of Abnormal Psychology © 2009 American Psychological Association

2009, Vol. 118, No. 3, 587–597 0021-843X/09/$12.00 DOI: 10.1037/a0016481

An 8-Year Longitudinal Study of the Natural History of Threshold,

Subthreshold, and Partial Eating Disorders From a Community
Sample of Adolescents
Eric Stice, C. Nathan Marti, Heather Shaw, and Maryanne Jaconis
University of Texas at Austin

The authors examined the natural history of threshold, subthreshold, and partial eating disorders in a
community sample of 496 adolescent girls who completed annual diagnostic interviews over an 8-year
period. Lifetime prevalence by age 20 years was 0.6% and 0.6% for threshold and subthreshold anorexia
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

nervosa (AN), 1.6% and 6.1% for threshold and subthreshold bulimia nervosa (BN), 1.0% and 4.6% for
This document is copyrighted by the American Psychological Association or one of its allied publishers.

threshold and subthreshold binge-eating disorder (BED), and 4.4% for purging disorder (PD). Overall,
12% of adolescents experienced some form of eating disorder. Subthreshold BN and BED and threshold
PD were associated with elevated treatment, impairment, and distress. Peak age of onset was 17–18 years
for BN and BED and 18 –20 years for PD. Average episode duration in months was 3.9 for BN and BED
and 5.1 for PD. One-year recovery rates ranged from 91% to 96%. Relapse rates were 41% for BN, 33%
for BED, and 5% for PD. For BN and BED, subthreshold cases often progressed to threshold cases and
diagnostic crossover was most likely for these disorders. Results suggest that subthreshold eating
disorders are more prevalent than threshold eating disorders and are associated with marked impairment.

Keywords: eating disorders, incidence, duration, recovery, diagnostic crossover

Studies of clinical samples indicate that eating disorders, includ- gression, and diagnostic crossover for a broad range of threshold,
ing anorexia nervosa (AN), bulimia nervosa (BN), and the provi- subthreshold, and partial eating disorders during this developmen-
sional diagnosis of binge-eating disorder (BED; American Psychi- tal period. We use the term subthreshold eating disorders to refer
atric Association, 1994), are associated with functional to individuals who experience all of the symptoms of a particular
impairment, emotional distress, psychiatric comorbidity, a chronic eating disorder but who endorse subthreshold levels of one or more
course, and elevated mortality risk (Grilo et al., 2003; Herzog et symptoms (e.g., only report engaging in binge eating and compen-
al., 2000; Keel, Mitchell, Miller, Davis, & Crow, 1999; Strober, satory behaviors once a week). We use the term partial eating
Freeman, & Morrell, 1997). Studies of community samples also disorders to refer to individuals who report only a subset of the
indicate that these eating disorders are associated with impairment, symptoms of a particular disorder (e.g., only report compensatory
a chronic course, and increased risk for subsequent psychiatric behaviors but not binge eating or vice versa). The lack of such data
problems and obesity (Cachelin et al., 1999; Fairburn, Cooper, for subthreshold and partial eating disorders, such as purging
Doll, Norman, & O’Connor, 2000; Hudson, Hiripi, Pope, & disorder (PD), is concerning because most adolescents who present
Kessler, 2007; Johnson, Cohen, Kotler, Kasen, & Brook, 2002; for eating disorder treatment do not satisfy criteria for Diagnostic
Patton, Coffey, Carlin, Sanci, & Sawyer, 2008; Råstam, Gillberg, and Statistical Manual of Mental Disorders (4th ed.; DSM–IV;
& Wentz, 2003). American Psychiatric Association, 1994) AN or BN (Fairburn &
Relatively fewer studies have characterized the natural history Harrison, 2003; Fisher, Schneider, Burns, Symons, & Mandel,
of the spectrum of eating disordered behavior in adolescents using 2001; Herzog, Hopkins, & Burns, 1993; Williamson, Gleaves, &
a prospective design. It is important to characterize the prevalence, Savin, 1992). Subthreshold and partial eating disorders are asso-
incidence, duration, recovery rates, relapse rates, diagnostic pro- ciated with functional impairment, distress, suicidal attempts, med-
ical complications, and increased risk for current and future psy-
chiatric and medical problems (Crow, Agras, Halmi, Mitchell, &
Eric Stice, C. Nathan Marti, Heather Shaw, and Maryanne Jaconis, Kraemer, 2002; Garfinkel et al., 1995; Keel, Haedt, & Edler, 2005;
Department of Psychology, University of Texas at Austin. Milos, Spindler, Schnyder, & Fairburn, 2005; Mond et al., 2006;
This study was supported by Career Award MH01708 and Research Stice, Marti, Spoor, Presnell, & Shaw, 2008; Striegel-Moore,
Grant MH/DK61957 from the National Institutes of Health. Thanks go to Seeley, & Lewinsohn, 2003). Scholars have called for further
project research assistants Sarah Kate Bearman, Cara Bohan, Emily research on subthreshold and partial eating disorders because they
Burton, Melissa Fisher, Lisa Groesz, Jenn Tristan, Natalie McKee, Kather-
represent examples of eating disorders not otherwise specified
ine Presnell, and Katy Whitenton; a multitude of undergraduate volunteers;
(EDNOS) when associated with functional impairment and dis-
the Austin Independent School District; and the participants who made this
study possible. tress (Fairburn & Harrison, 2003; Keel, 2007; Wilson, Becker, &
Correspondence concerning this article should be addressed to Eric Heffernan, 2003).
Stice, who is now at Oregon Research Institute, 1715 Franklin Boulevard, Epidemiological studies using diagnostic interviews indicate
Eugene, OR 97403. E-mail: [email protected] that the lifetime prevalence among women ranged from 0.9 to

587
 588 STICE, MARTI, SHAW, AND JACONIS

2.0% for AN, 1.1 to 4.6% for BN, 0.2 to 3.5% for BED, and 1.1 Few prospective studies of community-recruited samples report
to 5.3% for PD (Favaro, Ferrara, & Santonastaso, 2003; Hudson et the annual incidence of threshold, subthreshold, and partial eating
al., 2007; Lewinsohn, Striegel-Moore, & Seeley, 2000; Wade, disorders or the peak periods of risk for onset of these conditions.
Bergin, Tiggemann, Bulik, & Fairburn, 2006; Woodside et al., Virtually all that is known about typical ages of onset is based on
2001). The lifetime prevalence of partial AN has ranged from 2.4 retrospective data. For instance, on the basis of retrospective
to 3.7% (Favaro et al., 2003; Lewinsohn et al., 2000; Patton et al., reports from a community sample of young adults, Lewinsohn et
2008; Wade et al., 2006). However, partial AN definitions were al. (2000) found that the peak period of risk for onset of AN and
heterogeneous, including individuals who (a) endorse the criterion BN was between the ages of 16 and 17 years. In contrast, Hudson
for a low body weight but who endorse only one of the other three et al. (2007) found a mean age of onset of 19 years for AN, 20
AN symptoms, (b) endorse all AN symptoms except amenorrhea, years for BN, and 25 years for BED on the basis of retrospective
(c) endorse all AN symptoms except low body weight, or (d) data from community-recruited adults. An improved understand-
endorse only two of the four AN symptoms. The lifetime preva- ing of the peak periods of risk for onset of eating disorders is vital
lence of partial BN has ranged from 2.5% to 6.0% (Favaro et al., for the optimal timing of risk factor studies and preventive inter-
2003; Lewinsohn et al., 2000). The definitions of partial BN were ventions. The data may also advance etiologic models for these
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

also heterogeneous, including those who (a) endorse twice weekly disorders by implicating a role of certain developmental transitions
binge eating for at least 3 months but only one other BN symptom (e.g., menarche or school transitions). Thus, our second aim in this
(e.g., not reporting compensatory behaviors and weight or shape study was to investigate the annual incidence of threshold, sub-
overvaluation), (b) endorse all BN symptoms except binge eating, threshold, and partial eating disorders from early adolescence to
(c) endorse all BN symptoms except that binge eating and com- young adulthood using prospective data and with a focus on
pensatory behaviors occurred less than twice weekly, (d) endorse identifying the peak period of risk for onset of these conditions.
only compensatory behaviors, (e) endorse only binge eating, or (f) Most information on the typical duration of eating disorders,
endorse only two of the three BN symptoms. Such heterogeneity is recovery rates, and relapse rates is also based on retrospective data
concerning because in some studies (e.g., Favaro et al., 2003), (Hudson et al., 2007). Retrospective reports from community
BED and PD would have been classified as partial BN. It may not samples indicate that the mean duration of illness is between 1.7
be ideal to mix together individuals who endorse only binge eating, and 5.7 years for AN, 5.8 and 8.3 years for BN, and 8.1 and 14.4
individuals who endorse only compensatory behaviors, individuals years for BED (Hudson et al., 2007; Pope et al., 2006; Råstam et
with subthreshold levels of all BN symptoms, and individuals al., 2003). Retrospective reports from clinical samples indicate that
whose responses involve other permutations. Although it is an the mean duration of illness is between 9.3 and 14.7 years for AN,
empirical question, we think there may be value in separating 7.7 and 11.7 years for BN, and 14.4 years for EDNOS (Herzog et
partial variants of BN, including BED and PD, from subthreshold al., 1999; Milos et al., 2005). The studies involving clinical versus
cases and in requiring subthreshold cases to endorse each symptom community samples probably found longer illness duration be-
domain, even if some are below diagnostic threshold, as this might cause duration of illness, as well as illness severity and functional
result in more homogeneous groupings that may better predict impairment, predict treatment seeking (Keel et al., 2002). Yet, the
prognosis and treatment response. Although studies have provided wide range in estimates suggests that distinct sampling biases or
data on the lifetime prevalence of partial eating disorders (e.g., differences in recovery definitions influence the estimates or that
BED and PD), studies have not reported the prevalence of sub- retrospective reports have questionable validity. Pope et al. (2006)
threshold disorders. noted that prospective studies of representative community sam-
Further, most of the lifetime prevalence estimates are based on ples are needed to provide firmer estimates of episode duration.
retrospectively reported data from studies that required partici- Some prospective studies have investigated relapse and recov-
pants to report on their lifetime up to the point of study entry, ery rates for patient populations. Herzog et al. (1999) found that
raising concerns about the veracity of the diagnoses (Hudson et al., only 6% of AN patients and 41% of BN patients showed full
2007). Thus, our first aim in the present study was to document the recovery 1 year after initiating treatment. By 7.5-year follow-up,
lifetime prevalence of threshold, subthreshold, and partial eating 34% of AN patients and 74% of BN patients showed full recovery,
disorders by age 20 years using structured diagnostic interviews although 40% of AN patients and 35% of BN patients relapsed
administered annually over 8 years in a community sample. We after recovery during this follow-up. Fichter and Quadflieg (2007)
distinguished subthreshold AN and BN from BED and PD. As found that 44% of AN patients, 57% of BN patients, and 65% of
well, we used definitions of subthreshold AN, BN, and BED that BED patients showed recovery 2 years after initiating treatment,
required that participants endorse all DSM–IV symptoms for those with these percentages rising to 54%, 70%, and 78%, respectively,
disorders, even if some or all of the symptoms were below diag- by 6-year follow-up. However, clinical samples may provide un-
nostic thresholds. We selected frequency cutoffs for the specific representative estimates of natural recovery and relapse rates.
diagnoses that minimize the number of participants reporting dis- Fairburn et al. (2000) found, on the basis of repeated diagnostic
ordered eating behavior that would fall between diagnoses. We interviews, that over 50% of community-recruited individuals with
selected minimum frequency criteria (e.g., for binge eating and BN recovered over a 5-year follow-up without treatment but that
compensatory behaviors) on the basis of an earlier study that nearly 50% of those who recovered from BN relapsed in the
suggested that even relatively low symptom frequencies, such as subsequent year. Over a 6-month follow-up of another community
twice per month, are associated with functional impairment and sample, Keel et al. (2005) found that only 4% of individuals with
mental health treatment (Spoor, Stice, Burton, & Bohon, 2007). BN and 8% of individuals with PD showed recovery. Thus, our
Most diagnostic criteria were developed a priori in an independent third aim in this study was to characterize the episode duration,
study (Stice et al., 2008). recovery rates, and relapse rates of threshold, subthreshold, and
 NATURAL HISTORY OF EATING PATHOLOGY 589

partial eating disorders using prospective data collected over an from 12 to 15 years of age (M ⫽ 13 years) and were in 7th or 8th
8-year period. grade at baseline. The sample included 2% Asian/Pacific Islanders,
Although most information about progression from subthresh- 7% African Americans, 68% Caucasians, 18% Hispanics, 1%
old to threshold levels of an eating disorder and crossover from Native Americans, and 4% who specified other/mixed racial her-
one eating disorder diagnosis to another comes from retrospec- itage, which was representative of the schools from which we
tively reported data from cross-sectional studies, a handful of sampled. Average parental education, a proxy for socioeconomic
prospective studies have addressed these questions. Patton, status, was 29% high school graduate or less, 23% some college,
Johnson-Sabine, Wood, Mann, and Wakeling (1990) found that 33% college graduate, and 15% graduate degree, which was rep-
only 11% of adolescent girls with a diagnosis of partial syndrome resentative of the city from which we sampled.
eating disorders in a large community sample showed onset of BN The study was described as an investigation of adolescent men-
over a 1-year follow-up. Lewinsohn et al. (2000) found that only tal and physical health. An active parental consent procedure was
28% of individuals from a community sample who showed onset used to recruit participants, wherein an informed consent letter
of BN reported a history of partial BN previously and that the 1 describing the study was sent to parents of eligible girls (a second
individual who reported onset of AN did not report a previous mailing was sent to nonresponders). This resulted in an average
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

history of partial AN. With regard to diagnostic crossover, Milos participation rate of 56%, which was similar to the rate for other
et al. (2005) investigated a large sample of individuals with various school-recruited samples that used active consent procedures and
eating disorders who were primarily from treatment settings; over structured interviews (e.g., 61% for Lewinsohn et al., 2000).
a 30-month follow-up, the most frequent crossovers were from BN Participants completed a structured interview assessing eating dis-
to EDNOS (27%), AN to EDNOS (20%), EDNOS to AN (13%), order symptoms and had their weight and height measured by
and AN to BN (9%). Fichter and Quadflieg (2007) found that over female assessors at baseline (T1) and at seven annual follow-ups
a 2-year follow-up of a large clinical sample, the most frequent (T2, T3, T4, T5, T6, T7, and T8). Female assessors with at least a
crossovers were from BED to EDNOS (18%), BED to BN (15%), bachelor’s degree in psychology conducted the interviews. They
AN to BN (11%), and BN to EDNOS (7%). Råstam et al. (2003) attended 24 hr of training, wherein they received instruction in
reported a 22% conversion rate from AN to BN over a 6-year structured interview skills, reviewed diagnostic criteria for relevant
follow-up in a community sample. Another prospective commu- DSM–IV disorders, observed simulated interviews, and role-played
nity study found that diagnostic crossover rates were lower over a interviews. Assessors had to demonstrate an interrater agreement
6-month follow-up, with only 4% of individuals with PD crossing (␬ ⬎ .80) with supervisors using tape-recorded interviews before
over to BN and only 8% of individuals with BN crossing over to collecting data. Assessments took place at schools, participants’
PD (Keel et al., 2005). Thus, our fourth aim in this study was to houses, or the research offices. Participants received a $15 gift
examine diagnostic progression from subthreshold to threshold certificate or a $25 cash payment for completing each assessment.
diagnoses of the same disorder and diagnostic crossover from one
eating disorder to another.
Measures
In sum, in this study, we examined the (a) lifetime prevalence
and incidence of threshold and subthreshold AN, BN, BED, and Eating pathology. The Eating Disorder Diagnostic Interview
PD by age 20 years; (b) peak periods of risk for onset of these (EDDI; Stice et al., 2008), a semi-structured interview adapted
disorders; (c) duration and relapse rates for these eating distur- from the Eating Disorder Examination (Fairburn & Cooper, 1993),
bances; and (d) progression from subthreshold to threshold diag- assessed DSM–IV criteria for AN, BN, and BED over the past 12
noses of the same disorder and crossover from one eating disorder months at each of the eight annual assessments. The EDDI does
to another. An additional aim was to test whether individuals with not assess subjective binge-eating episodes because the 2–7 day
subthreshold eating disorders and those with PD evidence impair- test–retest reliability for subjective binge-eating episodes (r ⫽
ment, as indexed by elevated treatment, functional impairment, .33) is much lower than is the case for objective binge-eating
and emotional distress. We addressed these questions with data episodes (r ⫽ .85; Rizvi, Peterson, Crow, & Agras, 2000).
from a prospective risk factor study that followed a community Responses to these items allowed us to determine whether
sample of 496 adolescents over an 8-year period from early ado- participants met criteria for threshold, subthreshold, or partial
lescence to young adulthood. We focused on this age range be- eating disorders at any time point using a computer algorithm.
cause data suggest that the peak period of risk for eating pathology In accordance with previous studies (Garfinkel et al., 1995; le
onset may occur during this time (Lewinsohn et al., 2000). We Grange et al., 2006; Stice et al., 2008), participants who en-
focused on females because they are approximately 10 times more dorsed symptoms from each domain for a particular eating
likely to develop eating pathology than are males (Wilson et al., disorder but who endorsed a subthreshold level on at least one
2003). We also improved on some of the limitations of past symptom were given subthreshold diagnoses. Table 1 provides
epidemiologic studies by using more sensitive structured diagnos- the diagnostic criteria for the various eating disorders. Exces-
tic interviews on an annual basis and by following participants sive exercise was defined as at least 1 hr of vigorous exercise or
over a longer developmental period. 2 hr of moderate exercise that was engaged in specifically to
compensate for a binge-eating episode. Fasting was defined as
Method complete abstinence from caloric intake (meals or snacks) for
approximately 24 hr or more for the purpose of weight control
Participants and Procedures (American Psychiatric Association, 1994).
Participants were 496 adolescent girls recruited from public and To assess test–retest reliability for this adapted interview, we
private middle schools in a large U.S. city. Participants ranged randomly selected a subset of 137 participants who were inter-
 590 STICE, MARTI, SHAW, AND JACONIS

Table 1
Diagnostic Criteria for Eating Disorders

Diagnosis Criteria

Subthreshold anorexia nervosa ● BMI between 90% and 85% of that expected for age and gender
● Definite fear of weight gain more than 25% of the days for at least 3 months
● Weight and shape were definitely an aspect of self-evaluation
● Missed one period in a 3-month period (unless on birth control)
Threshold anorexia nervosa ● BMI less than 85% of that expected for age and gender
● Definite fear of weight gain more than 50% of the days for at least 3 months
● Weight and shape were one of the main aspects of self-evaluation
● Missing menstrual cycles in a 3-month period (unless on birth control)
Subthreshold bulimia nervosa ● At least two uncontrollable binge-eating episodes (i.e., objective binge-eating episodes) per
month for at least 3 months
● At least two compensatory behavior episodes (i.e., self-induced vomiting, laxatives use,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

diuretic use, fasting, and excessive exercise to compensate for overeating) per month for
This document is copyrighted by the American Psychological Association or one of its allied publishers.

at least 3 months
● Weight and shape was definitely an aspect of self-evaluation
Threshold bulimia nervosa ● At least eight uncontrollable binge-eating episodes per month for at least 3 months
● At least eight compensatory behavior episodes per month for at least 3 months
● Weight and shape was definitely one of the main aspects of self-evaluation
Subthreshold binge-eating disorder ● At least two uncontrollable binge-eating episodes/days per month for at least 6 months
● Less than one compensatory behavior on average per month during this period
● Marked distress about binge eating
● Binge eating characterized by three or more of the following: rapid eating; eating until
uncomfortably full; eating large amounts when not physically hungry; eating alone
because of embarrassment; feeling disgusted, depressed, or guilty after overeating
Threshold binge-eating disorder ● At least eight uncontrollable binge-eating episodes/days per month for at least 6 months
● Less than one compensatory behavior on average per month during this period
● Marked distress about binge eating
● Binge eating characterized by three or more of the following: rapid eating; eating until
uncomfortably full; eating large amounts when not physically hungry; eating alone
because of embarrassment; feeling disgusted, depressed, or guilty after overeating
Purging disorder ● At least eight episodes of self-induced vomiting or diuretic/laxative use for weight control
purposes per month for at least 3 months
● Less than one uncontrollable binge-eating episode on average per month during this period
● Weight and shape was definitely an aspect of self-evaluation

Note. A diagnosis of threshold or subthreshold anorexia nervosa took precedence over threshold and subthreshold diagnosis of bulimia nervosa and
binge-eating disorder, and purging disorder. Although it was not possible with the definitions we used, a diagnosis of threshold or subthreshold bulimia
nervosa took precedence over threshold or subthreshold binge-eating disorder and purging disorder.

viewed by the assessors for this study and another study (Stice et al., 1998). We used age- and sex-adjusted BMI centiles from the
al., 2008) to be reinterviewed by the same assessor within a Centers for Disease Control (Faith, Saelens, Wilfley, & Allison,
1-week period, resulting in high test–retest reliability (␬ ⫽ .96) for 2001) to determine whether participants were underweight for AN
the full range of threshold, subthreshold, and partial eating disor- diagnoses.
der diagnoses examined herein. To assess the interrater agreement Functional impairment. Impairment in the family, peer group,
for the eating disorder diagnoses, we randomly selected a subset of romantic, and school spheres was measured with 17 items from the
149 participants who were interviewed by the assessors for these Social Adjustment Scale Self-Report for Youth (Weissman, Orv-
two studies to be reinterviewed by a second assessor blind to initial aschel, & Padian, 1980; rated on a scale of 1 ⫽ never to 5 ⫽
diagnoses, resulting in high interrater agreement (␬ ⫽ .86) for the always). The original scale has shown convergent validity with
full range of eating disorder diagnoses. The EDDI has also been clinician and collateral ratings (M r ⫽ .72), discriminates between
shown to be sensitive to detecting intervention effects and has psychiatric patients and controls, and detects treatment effects
shown predictive validity for future onset of depression in past (Weissman & Bothwell, 1976). The 17-item version has shown
studies (Burton & Stice, 2006; Seeley, Stice, & Rohde, 2009; Stice internal consistency (␣ ⫽ .77) and 1-week test–retest reliability
et al., 2008). (r ⫽ .83), and it detects treatment effects (Burton & Stice, 2006;
Body mass. The body mass index (BMI ⫽ kg/m2; Pietrobelli Stice et al., 2008).
et al., 1998) was used for AN diagnoses. Height was measured to Mental health service use. An item assessing the frequency of
the nearest millimeter using portable stadiometers. Weight was visits to mental health care providers (i.e., “How often have you
assessed to the nearest 0.1 kg using digital scales with participants seen a psychologist, psychiatrist or other counselor/therapist be-
wearing light indoor clothing without shoes or coats. Two mea- cause of mental health problems in the last 6 months?”) was
sures of height and weight were obtained and averaged. The BMI generated for this study. This item showed 1-year test–retest reli-
shows convergent validity (rs ⫽ .80 –.90) with direct measures of ability (r) of .89 and sensitivity to detecting intervention effects
body fat such as dual energy X-ray absorptiometry (Pietrobelli et (Stice, Shaw, Burton, & Wade, 2006).
 NATURAL HISTORY OF EATING PATHOLOGY 591

Emotional distress. An item from the depression module of a particular eating disorder (or any others) for at least a 1-month
the Schedule for Affective Disorders and Schizophrenia for period. Relapse was defined as meeting criteria for another episode
School-Age Children (Puig-Antich & Chambers, 1983), a semi- of an eating disorder, as defined in Table 1, after showing at least
structured diagnostic interview, assessed subjective distress (i.e., a 1-month recovery from the same eating disorder. We then
“In the last 12 months did you have a period of time when you felt characterized the frequency of progressing from the subthreshold
sad, bad, unhappy, empty, or like crying for most of the day nearly level of an eating disorder to a full threshold level of the same
every day?”). Participants were asked to report their level of disorder (diagnostic progression) and the frequency of converting
emotional distress on a month-to-month basis over the 12-month from one eating disorder category to another (diagnostic cross-
reporting window using response options ranging from 1 ⫽ not at over). Participants had to satisfy all criteria for the subthreshold
all to 4 ⫽ severe. The 1-week test–retest reliability for this disorder specified in Table 1 (including the duration criteria) and
emotional distress question was .81 in the randomly selected subsequently satisfy all criteria for the threshold disorder specified
subset of 137 participants who were interviewed twice by inde- in Table 1 (including the duration criteria) to be coded as showing
pendent assessors over a 1-week period, providing support for the diagnostic progression. Participants had to satisfy all criteria for
reliability of this question. This emotional distress question also one disorder specified in Table 1 (including the duration criteria)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

correlated positively with Buss and Plomin’s (1984) Emotionality and subsequently satisfy all criteria for another disorder specified
Scale (r ⫽ .43) and the 17-item version of Weissman et al.’s in Table 1 (including the duration criteria) to be coded as showing
(1980) Social Adjustment Scale (r ⫽ .46) and inversely with the diagnostic crossover. Participants who showed the same diagnostic
Rosenberg (1979) Self-Esteem Scale (r ⫽ ⫺.43), providing evi- progression or crossover (e.g., transitioned from subthreshold BN
dence for the construct validity of this question. to threshold BN) on multiple occasions were counted only once.

Overview of Statistical Analyses Results

We first report the lifetime prevalence of threshold and sub- Attrition
threshold AN, BN, BED, and PD, which reflects the number of
participants who met criteria at baseline and the number of par- Between 1% and 8% of participants did not provide data at the
ticipants who showed onset of these eating disorders during the various follow-up assessments, but 99% of participants provided
8-year follow-up (Hoek, 2006). We refer to this as lifetime prev- data at baseline and at least one additional assessment. Attrition
alence by age 20 years because this was the mean age of partici- analyses indicated that there were no significant relations between
pants at T8. We also report the cumulative 8-year incidence for any of the eating disorder diagnoses and having missing data for
each of these disorders, which reflects the number of participants one or more follow-up assessments, suggesting that attrition did
who showed onset during the 8-year follow-up but does not not introduce systematic bias.
include participants who met criteria at baseline (Hoek, 2006). We
next used mixed models to test whether individuals with sub- Prevalence and Incidence of Eating Disorders
threshold and partial eating disorders reported more mental health
treatment, functional impairment, and emotional distress than did The lifetime prevalence by age 20 years, the cumulative incidence
non-eating-disordered participants. Mixed models also tested of onset during the 8-year follow-up, and the annual prevalence (the
whether subthreshold cases differed from threshold cases on these number of participants that exhibited a disorder during each annual
three impairment criteria. Models included all available data from interval in the follow-up period) of threshold and subthreshold AN,
the eight waves of the study as the dependent variable and BN, BED, and PD are reported in Table 2. The lifetime prevalence by
contained a random intercept that accounted for person-level vari- age 20 years ranged from a low of 1% for threshold BED to a high of
ation; eating disorder classification was the only fixed factor in the 6.1% for subthreshold BN. Overall, 12% of adolescents experienced
models. Variance explained (VE) is reported for significant ef- some form of eating disorder. The incidence over the 8-year follow-up
fects, using a pseudo-R2 formula (Kreft & de Leeuw, 1998). We period was slightly lower than the age 20 years lifetime prevalence
used an alpha of .01 to reduce the risk of chance findings resulting because at the baseline assessment, 3 participants met criteria for
from multiple testing. subthreshold BN, 2 met criteria for threshold BN, and 2 met criteria
After establishing that subthreshold eating disorders were gen- for subthreshold BED.1
erally associated with impairment, we collapsed threshold and Among individuals with subthreshold BN, the mean number of
subthreshold cases for each eating disorder for subsequent analy- binge-eating episodes per month was 3.4 (SD ⫽ 2.0), the mean
ses to increase cell sizes and enhance generalizability. We then number of compensatory behavior episodes per month was 9.7
derived noncumulative hazard curves for onset of the eating dis-
orders for ages 14 to 20 years on the basis of annual incidence data
1
to determine the peak period of risk for onset of each disorder. When we omitted the amenorrhea criterion for AN, the lifetime prev-
Next, for each eating disorder, we reported the average episode alence by age 20 years increased from 3 to 9 for subthreshold AN and from
3 to 10 for threshold AN; the cumulative incidence over the 8-year
duration in months, the 1-year and 2-year recovery rates, and the
follow-up increased from 3 to 9 for subthreshold AN and from 3 to 10 for
relapse rate during the follow-up period. Cases that met criteria for threshold AN. When we require only a 3-month duration rather than a
recovery or relapse were counted only once, even if they met 6-month duration for BED, the lifetime prevalence by age 20 years re-
criteria for each transition multiple times during the follow-up. mained 23 for subthreshold BED and increased from 5 to 6 for threshold
Following Agras, Walsh, Fairburn, Wilson, and Kraemer (2000), BED; the cumulative incidence over the 8-year follow-up remained 21 for
we defined recovery as not satisfying criteria shown in Table 1 for subthreshold BED and increased from 5 to 6 for threshold BED.
 592 STICE, MARTI, SHAW, AND JACONIS

Table 2
Incidence and Prevalence Rates for Eating Disorders in a Sample of 496 Adolescent Females Followed Over an 8-Year Period

Lifetime Annual prevalence

prevalence by Cumulative incidence
Eating disorder age 20 years over 8-year follow-up T1 T2 T3 T4 T5 T6 T7 T8

Subthreshold AN
n 3 3 0 0 2 1 0 0 0 0
% 0.6 0.6
95% CI ⫺0.1, 0.13 ⫺0.1, 0.13
Threshold AN
n 3 3 0 0 0 0 1 1 1 1
% 0.6 0.6
95% CI ⫺0.1, 0.13 ⫺0.1, 0.13
Subthreshold BN
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

n 30 26 3 2 5 5 8 10 9 5
This document is copyrighted by the American Psychological Association or one of its allied publishers.

% 6.1 5.2
95% CI 4.0, 8.2 3.3, 7.1
Threshold BN
n 8 6 2 1 0 0 1 4 2 0
% 1.6 1.2
95% CI 0.5, 2.7 0.2, 2.2
Subthreshold BED
n 23 21 2 4 1 1 4 5 7 8
% 4.6 4.2
95% CI 2.8, 6.4 2.4, 6.0
Threshold BED
n 5 5 0 0 0 0 1 2 3 2
% 1.0 1.0
95% CI 0.1, 1.9 0.1, 1.9
Purging disorder
n 22 22 0 5 1 0 4 4 6 7
% 4.4 4.4
95% CI 2.6, 6.2 2.6, 6.2

Note. AN ⫽ anorexia nervosa; BN ⫽ bulimia nervosa; BED ⫽ binge-eating disorder; CI ⫽ confidence interval; T ⫽ Time. Eating disorder classifications
are not mutually exclusive across disorders or between subthreshold and threshold cases (e.g., a participant can be classified as being subthreshold BN at
one time point, then threshold at another time point). Lifetime prevalence reflects the number of participants who met criteria at baseline and those who
showed onset of these eating disorders during the 8-year follow-up. Cumulative 8-year incidence reflects the number of participants who showed onset
during the 8-year follow-up, excluding participants who met criteria at baseline. Annual prevalence reflects the number of participants who met criteria for
a disorder at any point during each annual assessment period.

(SD ⫽ 11.8), and the mean rating for weight and shape overvalu- F(1, 2142) ⫽ 9.14, p ⫽ .003, VE ⫽ .02; and emotional distress,
ation was 4 (SD ⫽ 1.3). Among those with subthreshold BED, the F(1, 3056) ⫽ 21.00, p ⬍ .001, VE ⫽ .06, than did controls.
mean number of binge-eating episodes per month was 2.6 (SD ⫽ Participants with PD showed more mental health treatment,
2.2) and the mean number of compensatory behavior episodes per F(1, 2130) ⫽ 6.79, p ⫽ .009, VE ⫽ .02, and functional
month was 0.11 (SD ⫽ 0.3). Among those with PD, the mean impairment, F(1, 2137) ⫽ 8.32, p ⫽ .004, VE ⫽ .02, but not
number of purging episodes (i.e., vomiting, laxative use, or di- emotional distress, F(1, 3051) ⫽ 5.72, p ⫽ .02, VE ⫽ .02, than
uretic use) per month was 18.9 (SD ⫽ 8.3) and the mean number do control participants per our alpha level of .01. The means
of compensatory behavior episodes (which include fasting and and standard deviations for these groups on the impairment
excessive exercise) per month was 22.8 (SD ⫽ 10.6); none re- variables are shown in Table 3. Participants with subthreshold
ported binge-eating episodes during their PD episode. We do not BN did not differ from participants with threshold BN in levels
report descriptive statistics for threshold or subthreshold AN cases of mental health treatment, F(1, 152) ⫽ 0.87, p ⫽ .35, VE ⫽
because there were only 6 of them. .003; functional impairment, F(1, 152) ⫽ 0.84, p ⫽ .36, VE ⫽
.003; and emotional distress, F(1, 216) ⫽ 0.49,
Impairment p ⫽ .48, VE ⫽ .002. Participants with subthreshold BED did
Participants with subthreshold BN showed more mental not differ from participants with threshold BED in levels of
health treatment, F(1, 2169) ⫽ 8.27, p ⫽ .004, VE ⫽ .02; mental health treatment, F(1, 114) ⫽ 0.15, p ⫽ .70, VE ⫽ .001;
functional impairment, F(1, 2175) ⫽ 11.12, p ⬍ .001, VE ⫽ functional impairment, F(1, 114) ⫽ 0.09, p ⫽ .77, VE ⫽ .000;
.03; and emotional distress, F(1, 3103) ⫽ 18.18, p ⬍ .001, and emotional distress, F(1, 162) ⫽ 0.01, p ⫽ .92, VE ⫽ .000.
VE ⫽ .05, than did control participants. Participants with sub- All of these effects accounted for less than 1% of the variance,
threshold BED showed more mental health treatment, F(1, suggesting that the absence of effects was not solely due to
2136) ⫽ 11.56, p ⬍ .001, VE ⫽ .03; functional impairment, insufficient sensitivity because of modest cell sizes.
 NATURAL HISTORY OF EATING PATHOLOGY 593

Table 3
Means and Standard Deviations for the Impairment Variables for the Eating Disorder Groups and Non-Eating-Disordered
Participants

Mental health treatment Functional impairment Emotional distress

Eating disorder M SD M SD M SD

Subthreshold BN 2.48 3.70 2.36 0.51 1.85 0.69

Threshold BN 3.67 4.71 2.56 0.52 1.82 0.54
Subthreshold BED 2.83 3.82 2.37 0.52 1.93 0.70
Threshold BED 3.60 4.26 2.39 0.31 2.08 0.59
Purging disorder 2.37 2.58 2.35 0.31 1.70 0.37
Control participants 1.21 2.11 2.13 0.35 1.47 0.48

Note. BN ⫽ bulimia nervosa; BED ⫽ binge-eating disorder. Mental health treatment reflects the frequency of visits to any type of mental health care
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

provider. Functional impairment was rates on a 5-point scale ranging from 1 ⫽ never to 5 ⫽ always, with higher scores reflecting more impairment.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Emotional distress was rating on a 4-point scale ranging from 1 ⫽ not at all to 4 ⫽ severe.

In light of the evidence that subthreshold participants consis- with threshold or subthreshold BED recovered within 1 year
tently differed from non-eating-disordered participants but did not (96%), and 21 of the 22 individuals with PD recovered in 1 year
differ from threshold participants, we combined subthreshold and (95%). These recovery rates increased to 100% for each diagnostic
threshold participants with BN and BED for subsequent analyses. category by 2-year follow-up.
After combining subthreshold and threshold cases, 32 participants Among the 32 participants who exhibited threshold or sub-
exhibited subthreshold or threshold BN, 24 participants exhibited threshold BN, 19 experienced only one episode, 7 experienced two
subthreshold or threshold BED, and 22 participants exhibited PD. episodes, 3 experienced three episodes, 1 experienced four epi-
These numbers are slightly different than the numbers reported in sodes, and 1 experienced seven episodes, producing an overall
Table 2 because some participants showed diagnostic progression relapse rate of 41%. Among the 24 participants who exhibited
from subthreshold to threshold eating disorders and some showed threshold or subthreshold BED, 16 experienced only one episode,
diagnostic crossover. 5 experienced two episodes, 1 experienced four episodes, and 2
experienced six episodes, producing an overall relapse rate of
Peak Period of Risk for Onset 33%. Twenty-one of the 22 participants who exhibited PD epi-
sodes had only one episode and 1 participant had two episodes,
The noncumulative hazard rates for each eating disorder are producing an overall relapse rate of 5%.
presented in Figure 1. We excluded cases that were symptomatic at
baseline as we could not determine age of onset for these participants. Diagnostic Progression
Excluding cases that exhibited eating disorders at baseline, 27 partic-
ipants showed onset of subthreshold or threshold BN, 22 participants Among the 30 participants diagnosed with subthreshold BN, 5
showed onset of subthreshold or threshold BED, and 22 participants (17%) showed subsequent onset of full threshold BN during
showed onset of PD. The plotted hazard rates suggest that risk of follow-up. Among the 23 participants with a diagnosis of sub-
BN onset increased between ages 15 and 17 years, at which point threshold BED, 3 (13%) showed subsequent onset of threshold
onset peaked and began declining. The BED hazard rates did not BED during follow-up. It should be noted that the number of
exhibit an obvious peak but instead suggest that risk of onset is subthreshold cases is less than the total number of cases because
relatively constant across adolescence. The hazard rate for PD is not all participants satisfied the criteria for subthreshold diagnoses
rare before age 18 years and was generally highest in late adoles- shown in Table 1 for the required 3- or 6-month period before
cence. satisfying criteria for threshold versions of the same disorder (e.g.,
they only reported symptom frequencies below threshold levels for
a 1-month period).
Episode Duration
The average episode duration was 4.2 (SD ⫽ 3.3) months for Diagnostic Crossover
threshold and subthreshold BN, 3.9 (SD ⫽ 2.3) months for thresh- Crossover from BN to BED occurred in 6 of the 32 BN cases
old and subthreshold BED, and 4.7 (SD ⫽ 2.7) months for PD. It (19%) and crossover from BN to PD occurred in 0 of the 32 BN
is interesting that the episode duration was not always shorter for cases (0%). Crossover from BED to BN occurred in 10 of the 24
subthreshold variants of these eating disorders (i.e., whereas the BED cases (42%) and crossover from BED to PD occurred in 1 of
average duration of threshold BN was 3.6 months [SD ⫽ 2.3], the the 24 BED cases (4%). Crossover from PD to BN occurred in 1
average episode duration for subthreshold BN was 4.2 months of the 22 PD cases (5%) and crossover from PD to BED occurred
[SD ⫽ 3.6]). in 2 of the 22 PD cases (9%).

Recovery and Relapse Discussion

Twenty-nine of the 32 participants with threshold or subthresh- Our first aim in this study was to investigate the lifetime
old BN recovered within 1 year (91%), 23 of the 24 participants prevalence of threshold and subthreshold AN, BN, BED, and PD
 594 STICE, MARTI, SHAW, AND JACONIS
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 1. Noncumulative hazard functions of bulimia nervosa, binge-eating disorder, and purging disorder
onset by age. The y-axis depicts the noncumulative hazard rate for each disorder, which reflects the annual risk
for onset of the condition (annual incidence). The 95% confidence intervals for the annual incidence data are
shown in gray.

by age 20 years. Results indicated that the prevalence rates were Australia from a cross-sectional epidemiologic study (Wade et al.,
0.6% and 0.6% for threshold and subthreshold AN, 1.6% and 6.1% 2006).
for threshold and subthreshold BN, 1.0% and 4.6% for threshold The present findings indicate that many more participants re-
and subthreshold BED, and 4.4% for PD. These findings are ceived a diagnosis for subthreshold levels of AN and BN and other
important because this is the first study to report lifetime preva- EDNOS diagnosis (BED and PD) in total (17.3%) than received a
lence of threshold, subthreshold, and partial eating disorders from diagnosis for DSM–IV AN and BN (2.2%). The fact that a similar
a community-recruited sample that completed annual diagnostic pattern has emerged in samples of young adult women (e.g.,
interviews during the entire adolescent period. Most previous Favaro et al., 2003; Lewinsohn et al., 2000) suggests that the stage
studies relied on retrospective reports or less frequent interviews. may be set for eating pathology in adolescence, underscoring the
It is reassuring that our lifetime prevalence estimates through age importance of early prevention and treatment intervention. The
20 years are in the range reported in past epidemiologic studies of evidence that subthreshold BN and other EDNOS eating disorder
young adults for threshold (0.7 to 2.0%) and partial (2.4 to 3.7%) diagnoses are associated with elevated functional impairment,
AN and threshold (1.2 to 4.6%) and partial (2.5 to 6.0%) BN emotional distress, and treatment seeking also highlights the im-
(Favaro et al., 2003; Kjelsas, Bjornstrom, & Gotestam, 2004; portance of early intervention. In total, 12% of the participants met
Lewinsohn et al., 2000; Patton et al., 2008). The prevalence criteria for one or more of these eating disorders during adoles-
estimates for PD and subthreshold BED are novel contributions, as cence, suggesting that many young women pass through a period
these are the first lifetime prevalence estimates from a U.S. sample of disordered eating.
involving diagnostic interviews, although our PD prevalence esti- Our second aim in this study was to examine the peak periods of
mate is similar to the lifetime prevalence rate of 5.3% for PD in risk for onset of the various eating disorders. Peak period of risk
 NATURAL HISTORY OF EATING PATHOLOGY 595

for onset was between 17 and 18 years for BN and BED and was Our fourth aim in this study was to characterize progression
between 18 and 20 years for PD. To our knowledge, this is the first from subthreshold to threshold diagnoses of the same eating dis-
study to report on the peak period of risk for onset of PD using order and crossover from one eating disorder to another. Between
prospective data. The evidence that eating pathology tends to 13 and 17% of subthreshold cases progressed to threshold cases for
emerge in mid- to late adolescence suggests that developmental BN and BED during the study period. The diagnostic progression
experiences that typically occur in this age period, such as height- rates for BN and BED are similar to the relatively low rates
ened importance placed on conforming to the thin ideal precipi- observed in two previous prospective studies (Lewinsohn et al.,
tated by more time spent with peers and dating partners, may 2000; Patton, Johnson-Sabine, Wood, Mann, & Wakeling, 1990).
increase risk for eating pathology. It is possible that appetitive Higher rates of diagnostic progression have emerged in treatment-
drive to consume high-fat and high-sugar foods increases during seeking samples: Herzog et al. (1993) found that approximately
adolescence, potentially because of some changing biological pro- 50% of treatment-seeking women with a partial eating disorder in
cess (e.g., hormonal influence) or from conditioning processes a small case series went on to develop a full threshold eating
during this period. disorder.
There was also evidence of diagnostic crossover from one
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Our third aim in this study was to examine the episode duration,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

recovery rates, and relapse rates for these eating disturbances. disorder to another in the present community-recruited sample.
Average episode duration in months was 4.2 for threshold and Crossover was most likely from BED to BN (42% of BED cases),
subthreshold BN, 3.9 for threshold and subthreshold BED, and 4.7 followed by BN to BED (19% of BN cases). The rates of crossover
for PD. Although these durations are considerably shorter than from PD to BN and BED were much lower, similar to crossover
average duration estimates from treatment-seeking samples (e.g., rates observed in the one other prospective study of a community
Agras et al., 2000; Milos et al., 2005), other studies that examined sample of PD cases (Keel et al., 2005). The crossover estimates
community samples found relatively shorter duration estimates from BN to BED and vice versa were higher than the rates
(e.g., Lewinsohn et al., 2000). Presumably, individuals with the observed in Striegel-Moore et al.’s (2001) cross-sectional study
most severe and chronic forms of eating disorders seek treatment, (11% and 29%, respectively), potentially because the latter study
whereas less severe eating pathology may resolve more quickly. focused only on full threshold cases and relied on retrospective
report. This moderate diagnostic crossover suggests some fluidity
Recovery rates were high in this sample. Specifically, 1-year
in diagnoses, which might be useful to factor into nosological
recovery rates were 91% for threshold and subthreshold BN, 96%
frameworks, such as the DSM–IV. Fairburn and Harrison (2003)
for threshold and subthreshold BED, and 95% for PD. These rates
argued that this fluidity implies a fundamental problem with the
are much higher than those reported in prospective studies of
current diagnostic system. Given that the most common crossover
treatment-seeking samples (e.g., Fichter & Quadflieg, 2007; Her-
was from BN to BED and vice versa, researchers in future studies
zog et al., 1999), probably due to the fact that individuals who seek
should test whether alternative diagnostic approaches—such as a
treatment show more extreme eating pathology. Yet, the fact that
broad binge-eating syndrome that subtypes individuals as to
Fairburn et al. (2000) found a 50% recovery rate over a 1-year
whether they show consistent, intermittent, or no regular use of
period in their prospective study of community-recruited individ-
compensatory behaviors— has greater predictive validity than the
uals with threshold BN suggests that the relatively high recovery
present BN versus BED distinction in terms of clinical course and
rates reported in the present study may have emerged because most response to treatment.
of the present participants reported subthreshold eating disorders.
Another factor that might have contributed to the lower recovery
rates reported in previous studies is that they used more conser- Limitations
vative definitions of recovery. For instance, Herzog et al. (1999) It is important to consider the limitations of this study. First, the
required the absence of symptoms or the presence of only residual descriptive statistics should be interpreted with caution because the
symptoms for a period of 8 consecutive weeks for full recovery prevalence and incidence of the studied disorders are relatively
(versus 4 weeks for the present article). Reviews have established low and the sample size for this risk factor study was small relative
that there is wide variation in definitions of recovery that impact to epidemiological studies. Yet data from smaller studies in which
descriptive data regarding recovery rates for eating disorders (Cou- participants complete multiple detailed diagnostic interviews over
turier & Lock, 2006; Keel, Mitchell, Davis, Fieselman, & Crow, a long period of time complement data provided by large cross-
2000). Of note, however, the 1-year recovery rates were identical sectional epidemiological studies that rely on retrospective reports.
for each eating disorder even if we required the absence of diag- Ideally, the present assessment-intensive design would be used
nostic criteria for at least a 3-month period. with larger sample sizes. Second, retrospective data suggest that
Relapse rates in the present study were 41% for threshold and the average age of onset for certain eating disorders, including
subthreshold BN, 33% for threshold and subthreshold BED, and BED and PD, occurs in the early 20s (Hudson et al., 2007;
5% for PD. These findings are novel because few studies have Striegel-Moore et al., 2001; Wade et al., 2006), implying that the
characterized the course of subthreshold eating pathology and PD. estimates regarding peak periods of risk, prevalence, recovery,
It was noteworthy that relapse rates were highest for BN and BED, relapse, diagnostic progression, and diagnostic crossover may be
which may imply that recurrent binge eating results in some biased low, particularly given that the peak age of onset for PD
fundamental change that increases risk for reemergence of binge was 20 years. It is interesting, however, that Patton et al. (2008)
eating, such as conditioning in which cues associated with previ- found that few new cases of threshold or partial AN or BN
ous binge-eating episodes trigger cravings that result in the re- emerged during young adulthood in a prospective study. It is also
emergence of this behavior (Jansen, 1998). possible that participants showing onset during adolescence may
 596 STICE, MARTI, SHAW, AND JACONIS

differ qualitatively from those showing onset during young adult- (2002). Full syndromal versus subthreshold anorexia nervosa, bulimia
hood. Third, the descriptive statistics should also be interpreted nervosa, and binge eating disorder: A multicenter study. International
with care given the provisional nature of the definitions for sub- Journal of Eating Disorders, 32, 309 –318.
threshold and partial eating disorders. Fourth, participants who met Fairburn, C. G., & Cooper, Z. (1993). The eating disorder examination
criteria for an eating disorder were given a referral and encouraged (12th ed.). In C. Fairburn & G. T. Wilson (Eds.), Binge eating: Nature,
assessment, and treatment (pp. 317–360). New York: Guilford Press.
to seek treatment, which may have affected certain descriptive
Fairburn, C. G., Cooper, Z., Doll, H. A., Norman, P. A., & O. Connor,
statistics reported here (e.g., treatment rates, mean duration, and M. E. (2000). The natural course of bulimia nervosa and binge eating
diagnostic crossover). Fifth, the moderate recruitment rate may disorder in young women. Archives of General Psychiatry, 57, 659 –
limit the generalizability of the findings because those who en- 665.
rolled may differ from those who did not with regard to eating Fairburn, C. G., & Harrison, P. J. (2003). Eating disorders. Lancet, 361,
pathology. Although our prevalence estimates for eating disorders 407– 416.
are within the range observed in larger epidemiologic samples, Faith, M., Saelens, B., Wilfley, D., & Allison, D. (2001). Behavioral
future studies should strive for higher recruitment rates. Finally, treatment of childhood and adolescent obesity: Current status, chal-
participants were required to report on their symptoms over the lenges, and future directions. In J. K. Thompson & L. Smolak (Eds.),
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Body image, eating disorders, and obesity in youth (pp. 313–340).

This document is copyrighted by the American Psychological Association or one of its allied publishers.

previous 12 months at each annual assessment, which may have

introduced error. Washington DC: American Psychological Association.
Favaro, A., Ferrara, S., & Santonastaso, P. (2003). The spectrum of eating
disorders in young women: A prevalence study in a general population
Clinical and Research Implications sample. Psychosomatic Medicine, 65, 701–708.
Fichter, M. M., & Quadflieg, N. (2007). Long-term stability of eating
Findings suggest that eating pathology is more common that
disorder diagnoses. International Journal of Eating Disorders, 40, S61–
previously suspected, affecting approximately 12% of adolescent
S66.
girls. Although these conditions were not associated with the Fisher, M., Schneider, M., Burns, J., Symons, H., & Mandel, F. S. (2001).
protracted course and high relapse rates suggested by some previ- Differences between adolescents and young adults at presentation to an
ous clinical samples, they were characterized by functional im- eating disorder program. Journal of Adolescent Health, 28, 222–227.
pairment, emotional distress, treatment seeking, diagnostic pro- Garfinkel, P. E., Lin, E., Goering, P., Spegg, C., Goldbloom, D. S.,
gression and crossover, and moderate relapse rates. These data Kennedy, S., et al. (1995). Bulimia nervosa in a Canadian community
suggest that prevention efforts should be given a priority, partic- sample: Prevalence and comparison of subgroups. American Journal of
ularly those that have been found to reduce risk for onset of this Psychiatry, 152, 1052–1058.
broader array of eating disorders. Findings also emphasize the Grilo, C. M., Sanislow, C. A., Shea, M. T., Skodol, A. E., Stout, R. L.,
importance of offering eating disorder prevention programs during Pagano, M. E., et al. (2003). The natural course of bulimia nervosa and
middle adolescence, rather than young adulthood, if the goal of eating disorder not otherwise specified is not influenced by personality
disorders. International Journal of Eating Disorders, 34, 319 –330.
reducing risk for onset of eating pathology is to be realized. The
Herzog, D. B., Dorer, D. J., Keel, P. K., Selwyn, S. E., Ekeblad, E. R.,
findings may also have etiologic implications. For instance,
Flores, A. T., et al. (1999). Recovery and relapse in anorexia nervosa and
the high diagnostic crossover between BN and BED implies that bulimia nervosa: A 7.5 year follow-up study. Journal of the American
there may be shared risk factors for these conditions but that PD Academy of Child and Adolescent Psychiatry, 38, 829 – 837.
may be caused by a qualitatively distinct set of etiologic processes. Herzog, D. B., Greenwood, D. N., Dorer, D. J., Flores, A. T., Ekeblad,
Finally, although these findings provide important data bearing on E. R., Richards, A., et al. (2000). Mortality in eating disorders: A
the natural course of eating pathology among adolescents in the descriptive study. International Journal of Eating Disorders, 28, 20 –26.
community, it will be vital for larger studies involving more Herzog, D. B., Hopkins, J., & Burns, C. D. (1993). A follow-up study of
representative samples to investigate this phenomenon. 33 subdiagnostic eating disordered women. International Journal of
Eating Disorders, 14, 261–267.
Hoek, H. W. (2006). Incidence, prevalence, and mortality of anorexia
References nervosa and other eating disorders. Current Opinion in Psychiatry, 19,
Agras, W. S., Walsh, B. T., Fairburn, C. G., Wilson, G. T., & Kraemer, 389 –394.
H. C. (2000). A multicenter comparison of cognitive– behavioral therapy Hudson, J., Hiripi, E., Pope, H., & Kessler, R. (2007). The prevalence and
and interpersonal therapy for bulimia nervosa. Archives of General correlates of eating disorders in the National Comorbidity Survey Rep-
Psychiatry, 57, 459 – 466. lication. Biological Psychiatry, 61, 348 –358.
American Psychiatric Association. (1994). Diagnostic and statistical man- Jansen, A. (1998). A learning model of binge eating: Cue reactivity and cue
ual of mental disorders (4th ed.). Washington, DC: Author. exposure. Behaviour Research and Therapy, 36, 257–272.
Burton, E., & Stice, E. (2006). Evaluation of a healthy-weight treatment Johnson, J. G., Cohen, P., Kotler, L., Kasen, S., & Brook, J. S. (2002).
program for bulimia nervosa: A preliminary randomized trial. Behaviour Psychiatric disorders associated with risk for the development of eating
Research & Therapy, 44, 1727–1738. disorders during adolescence and early adulthood. Journal of Consulting
Buss, A. H., & Plomin, R. (1984). Temperament: Early developing per- and Clinical Psychology, 70, 1119 –1128.
sonality traits. Hillsdale, NJ: Erlbaum. Keel, P. K. (2007). Purging disorder: Subthreshold variant or full-threshold
Cachelin, F. M., Striegel-Moore, R. H., Elder, K. A., Pike, K. M., Wilfley, eating disorder? International Journal of Eating Disorders, 40, S89 –
D. E., & Fairburn, C. G. (1999). Natural course of a community sample S94.
of women with binge eating disorder. International Journal of Eating Keel, P. K., Dorer, D. J., Eddy, K. T., Delinsky, S. S., Franko, D. L., Blais,
Disorders, 25, 45–54. M. A., et al. (2002). Predictors of treatment utilization among women
Couturier, J., & Lock, J. (2006). What is recovery in adolescent anorexia with anorexia and bulimia nervosa. American Journal of Psychiatry,
nervosa? International Journal of Eating Disorders, 39, 550 –555. 159, 140 –142.
Crow, S. J., Agras, W. S., Halmi, K., Mitchell, J. E., & Kraemer, H. C. Keel, P. K., Haedt, A., & Edler, C. (2005). Purging disorder: An ominous
 NATURAL HISTORY OF EATING PATHOLOGY 597

variant of bulimia nervosa? International Journal of Eating Disorders, Seeley, J., Stice, E., & Rohde, P. (2009). Screening for depression preven-
38, 191–199. tion: Identifying adolescent girls at high risk for future depression.
Keel, P. K., Mitchell, J. E., Davis, T. L., Fieselman, S., & Crow, S. J. Journal of Abnormal Psychology, 118, 161–170.
(2000). Impact of definitions on the description and prediction of bu- Spoor, S. T., Stice, E., Burton, D., & Bohon, C. (2007). Relations of
limia nervosa. International Journal of Eating Disorders, 28, 377–386. bulimic symptom frequency and intensity to psychosocial impairment
Keel, P. K., Mitchell, J. E., Miller, K. B., Davis, T. L., & Crow, S. J. and health care utilization: Results from a community-recruited sample.
(1999). Long-term outcome of bulimia nervosa. Archives of General International Journal of Eating Disorders, 40, 505–514.
Psychiatry, 56, 63– 69. Stice, E., Marti, N., Spoor, S., Presnell, K., & Shaw, H. (2008). Dissonance
Kjelsas, E., Bjornstrom, C., & Gotestam, K. G. (2004). Prevalence of and healthy weight eating disorder prevention programs: Long-term
eating disorders in female and male adolescents (14 –15 years). Eating effects from a randomized efficacy trial. Journal of Consulting and
Behaviors, 5, 13–25. Clinical Psychology, 76, 329 –340.
Kreft, I., & de Leeuw, J. (1998). Introducing multilevel modeling. London: Stice, E., Shaw, H., Burton, E., & Wade, E. (2006). Dissonance and healthy
Sage. weight eating disorder prevention programs: A randomized efficacy
le Grange, D., Binford, R. B., Peterson, C. B., Crow, S. J., Crosby, R. D., trial. Journal of Consulting and Clinical Psychology, 74, 263–275.
Klein, M. H., et al. (2006). DSM–IV threshold versus subthreshold Striegel-Moore, R. H., Cachelin, F. M., Dohm, F.-A., Pike, K. M., Wilfley,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

bulimia nervosa. International Journal of Eating Disorders, 39, 462– D. E., & Fairburn, C. G. (2001). Comparison of binge eating disorder
This document is copyrighted by the American Psychological Association or one of its allied publishers.

467. and bulimia nervosa in a community sample. International Journal of

Lewinsohn, P. M., Striegel-Moore, R. H., & Seeley, J. R. (2000). Epide- Eating Disorders, 29, 157–165.
miology and natural course of eating disorders in young women from Striegel-Moore, R. H., Seeley, J. R., & Lewinsohn, P. M. (2003). Psycho-
adolescence to young adulthood. Journal of the American Academy of social adjustment in young adulthood of women who experience an
Child and Adolescent Psychiatry, 39, 1284 –1292. eating disorder during adolescence. American Academy of Child and
Milos, G., Spindler, A., Schnyder, U., & Fairburn, C. G. (2005). Instability Adolescent Psychiatry, 42, 587–593.
of eating disorder diagnoses: Prospective study. British Journal of Psy- Strober, M., Freeman, R., & Morrell, W. (1997). The long-term course of
chiatry, 187, 573–578. severe anorexia nervosa in adolescents: Survival analysis of recovery,
Mond, J., Hay, P., Rodgers, B., Owen, C., Crosby, R., & Mitchell, J. relapse, and outcome predictors over 10 –15 years in a prospective study.
(2006). Use of extreme weight control behaviors with and without binge International Journal of Eating Disorders, 22, 339 –360.
eating in a community sample: Implications for the classification of Wade, T. D., Bergin, J. L., Tiggemann, M., Bulik, C. M., & Fairburn, C. G.
bulimic-type eating disorders. Psychosomatic Medicine, 39, 294 –302. (2006). Prevalence and long-term course of lifetime eating disorders in
Patton, G. C., Coffey, C., Carlin, J. B., Sanci, L., & Sawyer, S. (2008). an adult Australian twin cohort. Australian and New Zealand Journal of
Prognosis of adolescent partial syndromes of eating disorders. British Psychiatry, 40, 121–128.
Journal of Psychiatry, 192, 294 –299. Weissman, M., & Bothwell, S. (1976). Assessment of social adjustment by
Patton, G. C., Johnson-Sabine, E., Wood, K., Mann, A. H., & Wakeling, A. patient self-report. Archives of General Psychiatry, 33, 1111–1115.
(1990). Abnormal eating attitudes in London schoolgirls—A prospective Weissman, M. M., Orvaschel, H., & Padian, N. (1980). Children’s symp-
epidemiological study: Outcome at 12-month follow-up. Psychological tom and social functioning self-report scales comparison of mothers’ and
Medicine, 20, 383–394. children’s reports. Journal of Nervous and Mental Disease, 168, 736 –
Pietrobelli, A., Faith, M., Allison, D., Gallagher, D., Chiumello, G., & 740.
Heymsfield, S. (1998). Body mass index as a measure of adiposity Williamson, D. A., Gleaves, D. H., & Savin, S. S. (1992). Empirical
among children and adolescents: A validation study. Journal of Pediat- classification of eating disorder not otherwise specified: Support for
rics, 132, 204 –210. DSM–IV changes. Journal of Psychopathology and Behavioral Assess-
Pope, H. G., Lalonde, J. K., Pindyck, L. J., Walsh, T., Bulik, C. M., Crow, ment, 14, 201–216.
S. J., et al. (2006). Binge eating disorder: A stable syndrome. American Wilson, G. T., Becker, C. B., & Heffernan, K. (2003). Eating disorders. In
Journal of Psychiatry, 163, 2181–2183. E. J. Mash & R. A. Barkley (Eds.), Child psychopathology (2nd ed., pp.
Puig-Antich, J., & Chambers, W. J. (1983). Schedule for Affective Disor- 687–715). New York: Guilford Press.
ders and Schizophrenia for School-Age Children (6–18 years). Pitts- Woodside, D. B., Garfinkel, P. E., Lin, E., Goering, P., Kaplan, A. S.,
burgh, PA: Western Psychiatric Institute. Goldbloom, D. S., & Kennedy, S. H. (2001). Comparison of men with
Råstam, M., Gillberg, C., & Wentz, E. (2003). Outcome of teenage-onset full or partial eating disorders, men without eating disorders, and women
anorexia nervosa in a Swedish community-based sample. European with eating disorders in the community. American Journal of Psychiatry,
Child and Adolescent Psychiatry, 12(Suppl. 1), 78 –90. 158, 570 –574.
Rizvi, S. L., Peterson, C. B., Crow, S. J., & Agras, W. S. (2000). Test–
retest reliability of the Eating Disorder Examination. International Jour- Received May 27, 2008
nal of Eating Disorders, 28, 311–316. Revision received May 4, 2009
Rosenberg, M. (1979). Conceiving the self. New York: Basic Books. Accepted May 4, 2009 䡲