artigo-SI PT
artigo-SI PT
artigo-SI PT
Therapy
for Adolescent Anorexia Nervosa
JAMES LOCK, M.D., PH.D., W. STEWART AGRAS, M.D., SUSAN BRYSON, M.A.,
AND HELENA C. KRAEMER, PH.D.
ABSTRACT
Objective: Research suggests that family treatment for adolescents with anorexia nervosa may be effective. This study was
designed to determine the optimal length of such family therapy. Method: Eighty-six adolescents (12–18 years of age) diagnosed
with anorexia nervosa were allocated at random to either a short-term (10 sessions over 6 months) or long-term treatment (20
sessions over 12 months) and evaluated at the end of 1 year using the Eating Disorder Examination (EDE) between 1999 and
2002. Results: Although adequately powered to detect differences between treatment groups, an intent-to-treat analysis found
no significant differences between the short-term and long-term treatment groups. Although a nonsignificant finding does not
prove the null hypothesis, in no instance does the confidence interval on the effect size on the difference between the groups
approach a moderate .5 level. However, post hoc analyses suggest that subjects with severe eating-related obsessive-
compulsive features or who come from nonintact families respond better to long-term treatment. Conclusions: A short-term
course of family therapy appears to be as effective as a long-term course for adolescents with short-duration anorexia nervosa.
However, there is a suggestion that those with more severe eating-related obsessive-compulsive thinking and nonintact families
benefit from longer treatment. J. Am. Acad. Child Adolesc. Psychiatry, 2005;44(7):632–639. Key Words: anorexia nervosa, family
therapy.
Anorexia nervosa usually first presents in Behavioral Sciences, Stanford University School of Medicine, 401
Quarry Road, Stanford, CA 94305; e-mail: Jimlock@stanford.edu.
adolescence. Despite the severity of this 0890-8567/05/4407–06322005 by the American Academy of Child
condition both medically and psychologically, it and Adolescent Psychiatry.
is still unclear how to treat this disorder DOI: 10.1097/01.chi.0000161647.82775.0a
(American Academy of Pediatrics, 2003). ies, this superiority was maintained 5 years after
However, the few existing controlled studies treatment was completed (Eisler et al., 1997).
suggest that family therapy that initially Among the treatment trials employing
promotes parental control over refeeding may be family therapy for adolescents, treatment
effective for this age group (Dare and Eisler, intensity averaged 20 sessions (range 9–36),
1997). Results of two of these trials provide whereas treatment duration averaged
preliminary evidence that this form of family approximately 12 months (range 6–36)
treatment is superior to individual therapy for (Lock and Le Grange, 2001). Comparison of
adolescents with anorexia nervosa (Robin et al., patient outcomes among these studies
1999; Russell et al., 1987) Further, on follow-up suggests that there was little difference in
in one of these stud- outcome as a result of having more
treatment sessions or longer treatment.
However, no study has evaluated the
Accepted December 21, 2004. relationship between the dose of therapy and
From the Department of Psychiatry and Behavioral Sciences, outcome (Le Grange et al., 1992; Robin et
Stanford University School of Medicine, Stanford, CA.
Dr. Lock’s work on this project was supported by NIH Career al., 1999; Russell et al., 1987). To advance
Development Award MH01457. our understanding of the effectiveness of
Correspondence to Dr. James Lock, Department of Psychiatry and
family therapy for adolescents with anorexia family treatment for anorexia nervosa available at the
time to estimate effect size (Le Grange et al., 1992;
nervosa, we undertook a study to determine Robin et al., 1999; Russell et al., 1987). For change in
the optimal length of treatment. Specifically body mass index (BMI), the primary outcome, a
we wished to determine whether patients sample of 86 subjects yields 80% power (assuming a
conservative moderate effect size, a standardized
treated for only 6 months (short-term) would mean difference of 0.5, and a 5% twotailed test of
do as well at the end of 1 year as those who significance) to detect group differences (Kraemer and
were treated for the entire period (longterm). Thienemann, 1987).
Baseline characteristics of participants in the study are
We hypothesized that there would be no
summarized on Table 1. The mean BMI at the initial point
difference in outcome between the short- of identification of the subjects for the study was 16.0 ± 1.6
and long-term groups. kg/m2. The mean BMI at the point of randomization to
METHOD treatments was 17.1 ± 1.4 kg/m 2. This mean increase in
BMI before randomization was a result of weight gain in
those adolescents (30% [26 patients]) briefly hospitalized
Participants (average length of stay was 12.3 days) for acute medical
Eighty-six adolescents (between the ages of 12 and instability before starting treatment. In addition, 31 (36%)
18 years; 77 females and 9 males) were entered into had comorbid psychiatric illness at baseline. Twenty-one
the study. All subjects completed human subjects (24%) had major depression or dysthymia, 12 (14%) had an
informed consent processes as approved by the anxiety disorder including obsessivecompulsive disorder,
institutional review board at the study site. These and four (5%) had other psychiatric disorders. Of these
adolescents met DSM-IV criteria for anorexia nervosa subjects, 12 (14%) received medications during the trial
except that some partially weight restored participants (selective serotonin reuptake inhibitors and atypical
were entered and for postmenarchal females, those antipsychotics), five in long-term treatment and seven in
who had missed a minimum of one menstrual period short-term treatment. There were no significant differences
instead of the usual three required by DSM-IV at the pretreatment evaluation between treatment groups on
criteria. This sample represents the typical range of any baseline measures.
adolescents presenting for outpatient treatment for
anorexia nervosa. Measures
Participants were recruited by referral from
pediatricians and therapists to a specialty evaluation Participants were assessed pretreatment and at 6 and 12
clinic for child and adolescent eating disorders (Fig. months using the Eating Disorder Examination (EDE), an
1). Screening of all referrals to this clinic (n = 241) interview that measures height, weight, and the severity of
from September 1999 to April 2002 yielded 141 the characteristic psychopathology of eating disorders
eligible participants over a 32-month period, of whom (Cooper et al., 1989). The adult version of this interview
86 (61%) agreed to randomization. Subjects screened was used because we found that adolescent participants
for the study included many patients with other eating were able to understand and respond to questions as
disorders (e.g., bulimia nervosa), which accounts for formulated in this iteration (Passi et al., 2003).
the high proportion of potential subjects screened out
initially. Reasons for nonparticipation were (1)
distance to treatment site (27%), (2) preferred
individual treatment (22%), (3) did not want to
participate in research (6%), and (4) no reason given
for refusal (45%).
Our plan called for exclusion of participants with
severe physical health problems likely to affect weight
(e.g., diabetes mellitus) or psychiatric illnesses that
would interfere with treatment (e.g., psychosis) and
those who had failed family treatment using the model
employed in the study, but no potential participants
were excluded based on these criteria. Psychotherapy
in addition to that offered in the study protocol was
not permitted. However, psychotropic medications
used to treat common comorbid psychiatric illnesses
(e.g., depression, obsessive-compulsive disorder,
generalized anxiety disorder) were allowed.
The sample size for this study was developed using
data from the previously published controlled trials of
Fig. 1 Flow diagram of subject progress during the randomized clinical trial.
These physicians were blinded to the randomization Regardless of the treatment phase, each session
status of the participants. Medical visits were begins with a 10– 15 minute check-in with the
scheduled at intervals determined by these physicians adolescent and is followed by a 45- to 50-minute
and ranged from twiceweekly visits to monthly visits, meeting with the entire family. Occasional brief
depending on their assessment of medical acuity, with telephone calls (less than 10 minutes in duration) are
an overall average of approximately 35 brief contacts allowed when problems arise that require consultation
per patient during the year. Patients were followed by with the therapist.
the physician group throughout the 1-year study In the current study, the short-term treatment
period regardless of the treatment group assigned. A focused on the first and second phases of treatment,
nutritionist was also available for brief consultations although at least one session, usually the last,
with participants and their parents during these visits examined more general adolescent issues. On the
as a part of the overall medical service; however, no other hand, the long-term treatment group allowed for
structured program of nutritional counseling was more time in each treatment phase and in particular
provided for subjects in the study. more time to focus on general adolescent concerns
during the third phase.
Treatments
The family treatment used is based on treatment Data Analysis
developed by Dare and colleagues at the Maudsley The primary analysis was by intention-to-treat. A
Hospital in London (Dare and Eisler, 1997). The random regression model was used to compare the
treatment was manual based and piloted before response trajectories (based on the EDE outcome data)
initiating the controlled trial (Lock and Le Grange, of the two groups. This assumes a linear model for
2001; Lock et al., 2001). This form of family each subject’s trajectory, such that the intercept
treatment is highly focused on the behaviors and represents the baseline response (which, given
thoughts associated with anorexia nervosa and sees randomization, should not differentiate the two
the adolescent as incapacitated mostly in terms of his groups), and the slope represents the rate of response
or her eating disorder specifically in the inability to for each individual subject. This approach uses the
maintain an optimal weight for age and height. In the repeated measures for each subject to reduce the
first phase, the therapist strives to unite the parents in unreliability of the response measure and is helpful in
developing a consistent approach to refeeding. The dealing with missing data and dropout. Effect sizes
therapist explicitly disclaims the notion that the (ES) are reported using the mean difference between
parents have caused the eating problem and instead groups divided by the pooled within-group SD. In a
expresses sympathy for the parents’ plight in trying to post hoc exploratory analysis of possible moderators
find a way to help their child. In addition, the therapist of treatment outcome, we employed a linear
attempts to recruit the sibling subsystem to support regression model using the measures at 1 year as the
their affected sibling. Parents work out for themselves dependent measure and controlling for baseline
how best to refeed their child with anorexia nervosa values. Independent variables were centered (Kraemer
with the therapist’s ongoing support and consultation. et al., 2002). For the analysis of the use of
Once steady weight gain is evident and the family hospitalization during treatment, we used a
experiences relief that they are being effective in MannWhitney U (a nonparametric test) to account for
taking charge of the eating disorder, the second phase nonnormally distributed values associated with this
of treatment begins. During this phase, symptoms of variable.
anorexia nervosa remain central in the discussions, but
other issues that are perceived to interfere with the
parents in their task of ensuring steady weight gain RESULTS
can be brought forward (e.g., eating at school or at
parties, level of activity, sports participation). Of the 86 participants, nine (10%) did not
When the patient achieves a stable weight and there
are no struggles with eating, the third phase of
complete treatment (defined before the start of
treatment is initiated. The goal of this phase is to study as attending 80% of assigned treatment
identify and briefly address other adolescent concerns sessions, i.e., eight of 10 sessions in the short-
that have been ignored because of the development of term treatment group and 16 of 20 in the long-
anorexia nervosa. Depending on the age of
development, this entails, among other things, term treatment group). Two (4%) dropped out of
working toward increased personal autonomy for the the short-term allocation and seven (16%)
adolescent, more appropriate family boundaries, and dropped out of the long-term allocation. The
the need for the parents to reorganize their life
together as their children mature. reasons for dropping out included a perception
by the parents that the treatment offered was not two outcomes, EDE eating and CBCL
effective (n = 1), a wish for a different form of Internalizing subscale, for which the long-term
treatment (n = 1), living too far away (n = 1), treatment has results better than those of the
need for other psychiatric treatment (n = 1), and short-term treatment (indicated by positive signs
participant refusal (n = 1). In the remaining on the ES of the slope, the standardized mean
cases (n = 4), the reason for dropout was difference between the slopes in the long-term
unknown because the participant could not be versus short-term treatments).
contacted. In addition, eight treatment Clearly, a nonsignificant difference does not
completers (three from the long-term treatment prove the null hypothesis, even when the study
group and five from the short-term group) did is adequately
not complete the final assessment (Fig. 1). powered. However, all the confidence intervals
for the ES indicate that the true ES is unlikely to
Treatment Outcome be anywhere near moderate (d = 0.5) (Cohen,
Comparison of intent-to-treat outcomes 1988. Another method for examining ES, the
between the two groups can be found in Table 2 area under the curve (Grissom, 1994; McGraw
and Figs. 2 and 3. Despite the fact that multiple and Wong, 1992) uses the probability that a
testing increases the probability of a false long-term treatment subject has an outcome
positive, no statistically significant difference preferable to that of a short-term treatment
between the treatment groups was detected on subject. In this case, area under the curve
the primary outcomes (BMI, EDE) or on any of calculations (with a null value of 50%) is at best
the secondary outcomes. Indeed there were only 53% for EDE Eating Concerns. The number
TABLE 2
Comparison of Long-Term and Short-Term Treatments
Long-Term Treatment Short-Term Treatment
(N = 42) Mean (SD ) (N = 44) Mean (SD )
Effect Size of
12 mo 6 mo Slope (95%
(End of (End of Confidence
Baseline 6 mo Treatment) Baseline Treatment) 12 mo Interval) AUC
Eating Disorder
Examination: Eating
Concerns Subscale 1.04 (1.33) 0.75 (1.00) 0.52 (0.83) 1.35 (1.13) 0.86 (1.01) 0.71 (0.92) +0.09 (–0.33 to +0.51) 53%
Eating Disorder Examination:
Restraint Subscale 2.64 (1.96) 1.64 (1.70) 1.42 (1.63) 2.76 (1.97) 1.84 (1.77) 1.62 (1.80) –0.08 (–0.51 to +0.34) 48%
Eating Disorder Examinations:
Shape Concerns
Subscale 2.41 (1.67) 1.96 (1.55) 1.76 (1.69) 2.61 (1.73) 2.25 (1.63) 2.08 (1.70) –0.12 (–0.54 to +0.31) 47%
Eating Disorder
Examinations: Weight
Concerns Subscale 1.96 (1.52) 1.62 (1.48) 1.39 (1.44) 2.32 (1.51) 2.01 (1.50) 1.97 (1.60) –0.11 (–0.53 to +0.31) 47%
Body mass index 17.3 (1.5)19.0 (1.8)19.5 (2.1)17.0 (1.3)19.0 (2.3)19.5 (2.2)–0.26 (–0.68 to +0.17) 43% Weight (kg) 46.7
(7.2) 51.4 (7.5)53.2 (8.0)44.6 (5.5)50.6 (8.1)52.0 (7.6)–0.21 (–0.63 to +0.22) 44%
Youth Self-Report
Externalizing Scale 9.2 (6.5) 8.6 (6.7) 8.5 (6.0) 9.4 (6.6) 9.5 (6.8) 9.6 (6.6) –0.19 (–0.62 to +0.25) 45%
Internalizing Scale 16.6 (11.7) 14.6 (10.7) 14.0 (10.8) 19.2 (12.3) 19.6 (12.6) 18.2 (12.2) –0.15 (–0.58 to +0.29) 46%
Total score 40.6 (25.6) 36.1 (23.7) 35.7 (23.1) 45.0 (27.7) 46.1 (28.3) 44.9 (27.6) –0.24 (–0.67 to +0.20) 43% Child
Behavior Checklist
Externalizing Scale 7.71 (6.6) 6.05 (5.8) 6.0 (5.2) 8.5 (6.3) 7.3 (5.5) 7.3 (5.9) –0.04 (–0.47 to +0.39) 49%
Internalizing Scale 15.6 (10.5) 12.6 (9.0) 11.2 (9.0) 18.7 (10.5) 15.7 (9.5) 13.8 (9.8) +0.04 (–0.40 to +0.47) 51%
Total Score 36.4 (25.4) 28.3 (17.7) 26.1 (19.1) 41.2 (21.3) 35.5 (20.7) 32.3 (22.6) –0.06 (–0.49 to +0.37) 48% Yale-
637 Brown Cornell Eating Disorder Scale: J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 44:7, JULY 2005
total score 12.2 (8.4) 8.8 (6.6) 6.4 (6.4) 13.4 (7.9) 10.9 (9.7) 9.2 (9.6) –0.28 (–0.70 to +0.15) 42%
AUC = area under the curve (probability that a long-term subject will have better outcome than a short-term subject).
Means and SD are presented. Significance tests and effect sizes compare the slopes in the two groups.
SHORT- AND LONG-TERM FAMILY THERAPY FOR ANOREXIA
Fig. 3 Change in global eating disorder examination by treatment group. Fig. 5 Change in global eating disorder examination by family status.
LOCK ET AL.
outcome measures (BMI and EDE) were education of the parents were on average higher
independently assessed by interviewers trained than expected in the general community,
and experienced in its use. Therefore, it seems although similar to many samples of participants
likely that both the therapy and the assessment with anorexia nervosa. Additional variables that
of eating-disordered psychopathological might have played a role in the overall outcome
processes were carried out with high fidelity, include the availability of medical
lending credibility to the results obtained. hospitalization for acutely ill patients both
The results of the study suggest that many before and during the study, use of medications,
adolescents with anorexia nervosa can be as and expectation effects of participation in a
effectively treated with a relatively short-term treatment study. It should also be emphasized
family therapy as with long-term family therapy. that treatments provided in this study were
It appears, however, that participants with more supported by an excellent medical staff with
severe and persistent eating-related obsessive- expertise in the treatment of adolescent anorexia
compulsive thinking or those who come from nervosa.
nonintact families did significantly better in
long-term treatment. Both of these potential
Clinical Implications
moderators make clinical sense. Those with the
highest degree of obsessional concern and most Overall, the results of this study suggest
intractable compulsive behaviors are more that for adolescents with anorexia nervosa, a
challenging for parents to redirect and thereby short-term treatment with family therapy is
might take longer to change. Intact families have as effective as long-term family therapy for
an advantage in taking on the tasks of changing the majority of such patients when their
behaviors related to anorexia nervosa because outcomes are assessed at the end of 1 year.
both parents are more likely to be available and A follow-up study to determine whether the
invested similarly in their child’s recovery. effects of treatment were maintained would
Because family-based treatment for anorexia be a useful future study. However, although
nervosa depends heavily on this parental there is no clear definition of remission or
resource, any compromise in it would likely lead recovery in anorexia nervosa, using BMI of
to the need for additional assistance. Thus, greater than 17.5 alone (the DSM-IV
longer term treatment might be expected. These diagnostic threshold for anorexia nervosa),
findings provide preliminary evidence to guide 96% of the sample are remitted at the end of
therapists when selecting patients for long- or treatment, whereas using a stricter criterion
short-term treatment. of a BMI of 20 and a global EDE score
within 2 SDs of normal, 67% would be
considered remitted.
Limitations
Clinicians interested in using family-
There are important qualifications relevant to based therapy for anorexia nervosa should
the current study that may limit its consider the possibility that a short-term
generalizability. The study sample consisted of course will benefit the patients; however, if
participants who agreed to family treatment there are exceptionally high levels of eating-
from referrals to a tertiary treatment center for related obsessionality and compulsiveness or
eating disorders. The conclusions therefore may the patient comes from a single-parent or
pertain to help-seeking families willing and able divorced family, it is probably more likely
to participate in family-based therapy. The that a longer course will be required. The
socioeconomic status of the sample and increase in the EDE scores for this subgroup
(Fig. 5) may portent future relapse. This American Academy of Pediatrics (2003), Identifying and
treating eating disorders: policy statement. Pediatrics
might suggest that therapists consider 111:204–211
alternative treatments (e.g., individual Cohen J (1988), Statistical Power Analysis for Behavioral
Science. Hillsdale, NJ: Erlbaum
treatment) in addition to more family Cook R (1995), The number needed to treat: a clinically
therapy. Longer term follow-up and useful measure of treatment effect. BMJ 310:452–454
Cooper Z, Cooper PJ, Fairburn C (1989), The validity of the
additional treatment studies may help to eating disorder examination and its subscales. Br J
shed light on this important subject. Psychiatry 154:807–812
Clinicians using this approach may not Dare C, Eisler I (1997), Family therapy for anorexia nervosa.
In: Handbook of Treatment for Eating Disorders, Garner
generally have the availability of medical D, Garfinkel D, eds. New York: Guilford, pp 307–324
experts to support their treatment as such a Dare C, Eisler I, Russell G, Treasure J, Dodge E (2001),
Psychological therapies for adults with anorexia nervosa:
resource is not always or commonly randomized controlled trial of outpatient treatments. Br J
available. At the same time, in the United Psychiatry 178:216–221
Efron B (1971), Forcing a sequential experiment to be
Kingdom, where this treatment originated, balanced. Biometrika 58:403–417
adolescents treated with this approach Eisler I, Dare C, Hodes M, Russell G, Dodge E, Le Grange D
generally are not provided with this level of (2000), Family therapy for adolescent anorexia nervosa:
the results of a controlled comparison oftwo family
medical support and appear to have similar interventions. J Child Psychol Psychiatry 41:727–736
outcomes (Eisler et al., 2000). It is also Eisler I, Dare C, Russell G, Szmukler G, Le Grange D,
Dodge E (1997), Family and individual therapy in
important to remember that in both arms of anorexia nervosa: a five-year followup. Arch Gen
the study, treatments were titrated upward Psychiatry 54:1025–1030
from a base of weekly sessions for Grissom R (1994), Probability of the superior outcome of
one treatment over another. J Appl Psychol 79:314–316
approximately 2 months at the start. Getting Kaufman J, Birmhaher B, Brent D et al. (1997), Schedule for
patients started in the right way appears to affective disorders and schizophrenia for school-age children–
present and lifetime version (KSADS-PL): initial reliability
be key in the success of this approach. The and validity data. J Am Acad Child Adolesc Psychiatry
decreased frequency of sessions later in 36:980–988
Kraemer H, Thienemann S (1987), How Many Subjects?
therapy may serve to boost and continue the Statistical Power Analysis in Research. Newbury Park, CA:
impact of the earlier sessions. Sage
Kraemer H, Wilson GT, Fairburn G, Agras WS (2002), Mediators
While these findings provide encouragement and moderators of treatment effects in randomized clinical
for the use of this form of treatment for the trials. Arch Gen Psychiatry 59:877–884
majority of such patients, albeit combined with Le Grange D, Eisler I, Dare C, Russell G (1992), Evaluation of
family treatments in adolescent anorexia nervosa: a pilot
hospitalization and psychotropic medications study.Int J Eat Disord 12:347–357
when needed, further research is needed to Lock J (1999), How clinical pathways can be useful: an example
of a clinical pathway for the treatment of anorexia nervosa in
determine definitively whether this type of adolescents Clin Child Psychol Psychiatry. 4:331–340
family therapy is superior to other forms of Lock J, Le Grange D (2001), Can family-based treatment of
anorexia nervosa be manualized? J Psychother Pract Res
treatment for adolescent anorexia nervosa. 10:253–261
Lock J, Le Grange D, Agras WS, Dare C (2001), Treatment
Manual for Anorexia Nervosa: A Family-Based Approach.
Disclosure: The authors have no financial New York: Guilford
relationships to disclose. Mazure S, Halmi C, Einhorn A (1995), The Yale-Brown-Cornell
Eating Disorder Scale: a new scale to assess eating disorder
symptomatology. Int J Eat Disord 18:237–245
Mazure S, Halmi S, Sunday S, Romano S, Einhorn A (1994), The
REFERENCES YaleBrown-Cornell Eating Disorder Scales: development, use,
reliability, and validity. J Psychiatr Res 28:425–445
Achenbach T (1991), Manual for the Child-Behavior McGraw K, Wong S (1992), A common language effect size
Checklist/4-18 and 1991 statistic. Psychol Bull 111:361–365
Profile. Burlington: University of Vermont, Department of Moos R (1974), The family environment scale, Form R. Palo Alto,
Psychiatry CA: Consulting Psychologists Press
Moos R, Moos B (1994), Family Environment Scale Manual, Palo
Alto, CA: Consulting Psychologists Press