Professional Documents
Culture Documents
Institutional Review Board at each site. After describing analysis for our investigation of moderators and non-
the study to the participating families, written informed specific predictors of early hospitalization. In this explor-
consent was obtained from the parents and an assent atory analysis we relaxed the significance level
from the adolescent. Details describing this studys pro- (alpha5.10) to allow detection of potential moderators in
tocol and primary outcomes are available elsewhere.4 this relatively small sample.15
For this study we examined the hospitalization data for
each of the participants together with relevant baseline
data. Medical and psychiatric hospitalization admission
and discharge decisions were made by clinicians blind to Results
the treatment randomization of participants. Medical
hospitalizations, the main type utilized in this study, Participants
were based on guidelines by the Society of Adolescent Data used in this study were collected from 158
Medicine for the inpatient treatment of adolescents with (78 in FBT, 80 in SyFT) adolescents (ages 12 to 18
anorexia nervosa.5 years of age) who met DSM-IVTR criteria for AN
(exclusive of the amenorrhea criteria). The mean
Measures age of participants was 15.3 (1.8) years, 89.2% were
The following measures were used in this study: female, the average duration of AN was 13.5 (13.9)
months. Full details of participant characteristics
1. Hospital days for individuals and by treatment includ- and the study protocol are available elsewhere.4 As
ing both medical and psychiatric hospitalizations. reported in the original article there were no differ-
2. Timing (by weeks of treatment) of hospital usage. Hos- ences between groups for any demographic or
pitalizations were divided into early (during the first 5- baseline variable. There were also no statistically
weeks of treatment) and late (after 5-weeks of treat- significant differences between treatments for per-
ment) based on examination of the data. cent ideal body weight, eating disorder symptoms,
3. Weight defined as percent mean body weight for age, or comorbid psychiatric disorders at either end of
height and gender using CDC norms.6 treatment or 1 year follow-up.
4. Attrition was defined as either dropping out of treat-
ment or the study.
Pattern of Hospitalization Use
Other secondary outcome measures used were the Figure 1 illustrates two distinct patterns of hos-
Eating Disorder Examination (EDE),7 Yale-Brown- pital use during treatment based on treatment
Cornell-Eating Disorder Scale (YBC-ED),8 Rosenberg assignment (FBT 5 12; SyFT 5 15). Most hospital-
Self-Esteem Scale (RSE),9 Beck Depression Inventory izations (73%) occurred in the first 5 weeks of treat-
(BDI),10 Child-YaleBrown Obsessive Compulsive Scale ment (early hospitalization). The median number
(CY-BOCS),11 Kiddie-Schedule for Affective Disorders in days for those hospitalized early for FBT was 7
and Schizophrenia KSADS),12 Multidimensional Perfec- (Interquartile Range 5 20) compared with 23 for
tions Scale (MPS)13 as described in the original article. SyFT (Interquartile Range 5 21), Mann-Whitney
For purposes of this study only baseline measures were U 5 28.5, p 5 0.018. While hospitalizations contin-
used as potential predictors or moderators of ued throughout treatment in SyFT, there were few
hospitalization. later hospitalizations (after 5 weeks of treatment)
for FBT. For FBT 2 patients (16.6% of early hospital-
Statistical Analysis izations) were rehospitalized after the first 5-weeks
Because the hospital data were not normally distrib- of treatment and for SyFT 6 patients (40%) were
uted, nonparametric statistics were used as appropriate. rehospitalized, accounting for a total of 6 rehospi-
Analyses were conducted in line with the intent-to-treat talizations for FBT and 13 for SyFT because of mul-
principle. The timing of hospitalization and use of hospi- tiple hospitalizations per patient. The total number
tal days between the two treatments was investigated by of hospital days used in SyFT was 655 and 369 in
plotting the variables and examining the effects on hos- FBT. As noted previously the median number
pitalization by treatment group. We also examined the of days per hospitalization was 21.0 for SyFT and
effect of hospitalization on dropout from treatment and 8.3 for FBT, Mann-Whitney U 5 51.0, p 5 0.02.4
on clinical outcome. In addition, we examined all base- Although not statistically significant, weight gain
line variables as potential moderators of treatment differed between groups during the first 4-weeks of
assignment effect on early hospitalization in line with treatment for those hospitalized early (9.0 lbs for
the McArthur approach.14 We used logistic regression FBT and 0.2 lbs for SyFT).
Effects of Early Hospitalization on Outcomes Type 2 error in the moderator analysis in this small
There was no effect of early hospitalization sample, we relaxed the Type 1 error rate (i.e.,
on treatment attrition or later hospitalization dur- change the alpha from 0.05 to 0.10). In this explora-
ing the treatment period. However, early hospitali- tory analysis, we identified two potential modera-
zation was a negative predictor (see Table 1) for tors of treatment effect: baseline BDI (p 5 0.094)
the percent weight change for both treatment and YBC-ED (p 5 0.078). The YBC-ED score has
groups. been previously identified as a moderator of treat-
ment effect in studies of FBT on recovery rates.3,16
Participant Characteristics Predictive of Early If these moderators were confirmed in a larger
Hospitalization sample of hospitalized patients, this would suggest
that FBT decreases early hospitalization specifically
In line with our original article,10 we examined
for individuals with higher baseline BDI and YBC-
16 baseline variables as potential moderators of
ED scores.
treatment effect on early hospitalization. None of
these variables were identified as moderators using
our a priori significance level of alpha 5 0.05 (two-
tailed), but six variables were found to be nonspe-
Discussion
cific predictors of early hospitalization (see Table 1 The current study extends findings from previous
for descriptive data). That is, regardless of the treat- studies related to FBT effects on hospitalization by
ment assignment status (i.e., FBT vs. SyFT), partici- illustrating the difference in timing of hospitaliza-
pants with lower baseline scores on the BDI, YBC- tion with FBT compared to SyST (see Fig. 1). When
ED. MPI, or RSE and lower baseline rates of com- hospitalization occurs in FBT, it occurs early then
pensatory behavior and comorbidity were less levels off; whereas in SyFT hospitalization rates not
likely to be hospitalized. Given the possibility of only occur early but continue to increase through-
out treatment. These differences in hospitalization
FIGURE 1 Cumulative hospital days over time and by treatment. patterns are likely the result of greater weight gain
FBT (family-based treatment) SyFT (systemic family therapy). [Color fig-
ure can be viewed in the online issue, which is available at wileyonli-
early in FBT4 compared to SyFT. Early weight gain
nelibrary.com.] during the first 5 weeks of treatment is protective
of later hospitalization because it decreases vulner-
ability to medical instability. Regardless of treat-
ment allocation, hospitalization during treatment
was predicted by the presence of obsessive-
compulsive symptoms related to eating (i.e., YBC-
ED), lower self-esteem, higher perfectionism,
depressive symptoms, and greater psychiatric co-
morbidity. Because the importance of identifying
moderators of treatment effect is high15 and the
rarity of even moderately large samples of adoles-
cents with AN in treatment studies, we relaxed our
significance level to alpha 5 0.10 in a highly explor-
atory hypothesis generating analysis. This yielded
two potential moderators: YBC-ED and BDI.
Importantly, higher scores on the YBC-ED has 3. Lock J, Le Grange D, Agras WS, Moye A, Bryson S, Jo BA. randomized clinical
trial comparing family based treatment to adolescent focused individual
been found to moderate clinical outcome (i.e.,
therapy for adolescents with anorexia nervosa. Arch Gen Psychiatry 2010;67:
recovery) in other studies of adolescent AN.3,16 10251032.
These exploratory findings are potentially impor- 4. Agras W, Lock J, Brandt H, Bryson S, Dodge E, Halmi K, et al. Comparison of
tant in developing hypotheses to be tested in future 2 family therapies for adolescent anorexia nervosa: A randomized parallel
confirmatory studies. trial. JAMA Psychiatry 2014;72:12791286.
5. Golden N, Katzman D, Kreipe R, Stevens S, Sawyer S, Rees J, et al. Eating dis-
There are significant limitations to the findings
orders in adolescents: Position paper of the Society for Adolescent Medici-
of this study. The sample size of hospitalized par- ne:Medical Indications for Hospitalization in an Adolescent with an Eating
ticipants is relatively small and data were collected Disorder. J Adolesc Health 2003;33:496503.
from hospital charts, discharge letters, and self- 6. Le Grange D, Doyle P, Swanson S, Ludwig K, Glunz C, Kreipe R. Calculation
reports rather than from a systematic data base. of expected body weight in adolescents with eating disorders. Pediatrics
2012;129:e438e446.
The sample used in the moderator analysis is likely
7. Cooper Z, Fairburn CG. The Eating Disorder Examination: A semi-structured
too small to rule out Type 2 error and therefore interview for the assessment of the specific psychopathology of eating disor-
there may be moderators which were not able to ders. Int J Eat Disord 1987;6:18.
be identified in this sample despite relaxing the sig- 8. Mazure S, Halmi CA, Einhorn A. The Yale-Brown-Cornell Eating Disorder
nificance level (Type 1 error rate) for this analysis. Scale: A new scale to asses eating disorder symptomatology. Int J Eat Disord
The results should be considered primarily applica- 1995;18:237245.
9. Rosenberg M. Conceiving the Self. New York: Basic Books; 1979.
ble to settings that use similar medical criteria for
10. Beck AT. Beck Depression Inventory. San Antonio, TX: The Psychological Cor-
hospitalization as those used in this study. None- poration; 1987.
theless, health planners looking to contain costs 11. Freeman J, Flessner C, Garcia A. The Childrens Yale Brown Obsessive Com-
while providing an effective treatment for adoles- pulsive Scale: Reliability, and validity for use among 5 to 8 year olds with
cents with AN should consider supporting the obessive compulsive disorder. J Abnorm Psychol 2011;39:877883.
12. Orvaschel H, Puig-Antich J, Chambers W, Tabrizi M, Johnson R. Retrospective
training and implementation of FBT for this
assessment of pre-pubertal major depression with the Kiddie-SADS-E. J Am
population. Acad Child Adolesc Psychiatry 1982;21:392
13. Frost R, Marten P, Lahart C, Rosenblate R. The dimensions of perfectionism.
Cognit Ther Res 1990;14:449468.
References 14. Kraemer H, Wilson GT, Fairburn CG, Agras WS. Mediators and moderators of
treatment effects in randomized clinical trials. Arch Gen Psychiatry 2002;59:
1. Striegel-Moore R, Leslie D, Petrill SA, Garvin V, Rosenheck RA. One-year use 877884.
and cost of inpatient and outpatient services among female and male 15. Kraemer H. Discovering, comparing, and combining moderators of treat-
patients with an eating disorder: Evidence from a national database of ment on outcome after randomized clinical trials: A parametric approach.
health insurance claims. Int J Eat Disord 2000;27:381389. Stat Med 2013;32:19641973.
2. Gowers S, Clark A, Roberts C, Griffiths A, Edwards V, Bryan C, et al. Clinical 16. Lock J, Agras WS, Bryson S, Kraemer H. A comparison of short- and long-
effectiveness of treatments for anorexia nervosa in adolescents. Br J Psy- term family therapy for adolescent anorexia nervosa. J Am Acad Child Ado-
chiatr 2007;191:427435. lesc Psychiatry 2005;44:632639.