Content uploaded by Anja Hilbert
Author content
All content in this area was uploaded by Anja Hilbert
Content may be subject to copyright.
Pretreatment and Process Predictors of Outcome in Interpersonal and
Cognitive Behavioral Psychotherapy for Binge Eating Disorder
Anja Hilbert
Philipps University of Marburg
Brian E. Saelens
Seattle Children’s Hospital and Regional Medical Center and the
University of Washington
Richard I. Stein
Washington University in St. Louis
Danyte S. Mockus
San Diego State University
R. Robinson Welch
Washington University in St. Louis
Georg E. Matt
San Diego State University
Denise E. Wilfley
Washington University in St. Louis
The present study examined pretreatment and process predictors of individual nonresponse to psychological
group treatment of binge eating disorder (BED). In a randomized trial, 162 overweight patients with BED
were treated with either group cognitive– behavioral therapy or group interpersonal psychotherapy. Treatment
nonresponse, which was defined as nonabstinence from binge eating, was assessed at posttreatment and at 1
year following treatment completion. Using 4 signal detection analyses, greater extent of interpersonal
problems prior to treatment or at midtreatment were identified as predictors of nonresponse, both at
posttreatment and at 1-year follow-up. Greater pretreatment and midtreatment concerns about shape and
weight, among those patients with low interpersonal problems, were predictive of posttreatment nonresponse.
Lower group cohesion during the early treatment phase predicted nonresponse at 1-year follow-up. Attention
to specific pre- or intreatment predictors could allow for targeted selection into differential or augmented care
and could thus improve response to group psychotherapy for BED.
Keywords: predictor, group psychotherapy, interpersonal psychotherapy, cognitive–behavioral therapy,
signal detection analysis
Binge eating and associated psychopathology in binge eating dis-
order (BED) can be substantially reduced through cognitive–
behavioral therapy (CBT) and through interpersonal psychotherapy
(IPT; National Institute for Clinical Excellence, 2004). Yet, 20%–
50% of patients fail to fully respond by treatment’s end, and effects
tend to wane in the long term (Wilson & Fairburn, 2002; Wonderlich,
de Zwaan, Mitchell, Peterson, & Crow, 2003). Establishing outcome
predictors could inform interventions and, thus, could prevent poor
response from patients with BED.
Higher initial binge eating (Loeb, Wilson, Gilbert, & Labouvie,
2000; Peterson et al., 2000) and more severe overeating problems
(Agras et al., 1995; Agras, Telch, Arnow, Eldredge, & Marnell, 1997)
appear to predict poorer posttreatment BED outcome. Evidence is
mixed as to whether initial specific eating disorder psychopathology,
general psychopathology, and self-esteem are related to outcome
(Agras et al., 1997; Carter & Fairburn, 1998; Loeb et al., 2000;
Peterson et al., 2000; Safer, Lively, Telch, & Agras, 2002), but BED
patients of the high negative affectivity subtype have particularly poor
treatment response (Loeb et al., 2000; Stice et al., 2001). Earlier age
of binge eating onset, binge eating preceding first dieting attempt
(Agras et al., 1995, 1997; Safer et al., 2002), and younger age when
receiving treatment (Agras et al., 1997) have been found to be related
to poor treatment response, whereas initial weight status is not pre-
dictive (Carter & Fairburn, 1998; Loeb et al., 2000).
Prediction of poor treatment outcome based on pretreatment
patient characteristics warrants clarification and replication, given
the few BED predictor studies and that inconsistent results are
likely related to a lack of statistical power and differences in
treatment modalities, length of follow-up, and definition and as-
Anja Hilbert, Department of Psychology, Philipps University of Mar-
burg, Marburg, Germany; Brian E. Saelens, Seattle Children’s Hospital and
Regional Medical Center and the University of Washington; Richard I.
Stein, Department of Internal Medicine, Washington University School of
Medicine; Danyte S. Mockus, Graduate School of Public Health, and
Georg E. Matt, Department of Psychology, San Diego State University; R.
Robinson Welch and Denise E. Wilfley, Department of Psychiatry, Wash-
ington University in St. Louis.
This research was supported by National Institute of Mental Health
Grants R29 MH51384, R29 MH138403, and K24 MH070446 to Denise E.
Wilfley and by German Ministry of Education and Research Grant
01GP0491 to Anja Hilbert. We are grateful to Helena C. Kraemer for her
statistical advice on signal detection analysis.
Correspondence concerning this article should be addressed to Anja
Hilbert, Philipps University of Marburg, Department of Psychology,
Gutenbergstrasse 18, D-35032 Marburg, Germany. E-mail: hilbert@
staff.uni-marburg.de
Journal of Consulting and Clinical Psychology Copyright 2007 by the American Psychological Association
2007, Vol. 75, No. 4, 645–651 0022-006X/07/$12.00 DOI: 10.1037/0022-006X.75.4.645
645
sessment of outcome across studies. Our focus in the present study
was, therefore, on examining as outcome predictors those patient
characteristics that were previously investigated and that are cen-
tral to CBT and to IPT treatment models (e.g., dietary restraint or
interpersonal problems, respectively) or to both. Within-treatment
processes may also prove predictive, as these factors predict treat-
ment outcome for bulimia nervosa (Agras et al., 2000; Fairburn,
Agras, Walsh, Wilson, & Stice, 2004; Loeb et al., 2005; Wilson,
Fairburn, Agras, Walsh, & Kraemer, 2002; Wilson et al., 1999).
Rapid reduction of binge eating predicts better posttreatment out-
come for individuals with BED who have undergone CBT (Grilo,
Masheb, & Wilson, 2006), but the predictive value of changes in
associated psychological symptoms, such as interpersonal prob-
lems, during the early treatment phase remains unclear. Nonspe-
cific process-related factors, such as therapeutic alliance or group
cohesion, have not been examined as outcome predictors for BED.
The present study examines pre- and intreatment factors as pre-
dictors of immediate and of long-term nonresponse in an ade-
quately powered, randomized-controlled trial of psychotherapy for
individuals with BED (Wilfley et al., 2002).
Method
Participants and Procedure
Participants were 162 overweight individuals with BED, who
were recruited for the treatment trial at two sites: New Haven (126
participants; 77.8%) and San Diego (36 participants; 22.2%).
Methods and design are detailed in the main outcome report
(Wilfley et al., 2002). Participants met diagnostic criteria for BED
according to the Diagnostic and Statistical Manual of Mental
Disorders (4th ed., text rev.; DSM–IV–TR;American Psychiatric
Association, 2000) and were randomized into either Group CBT or
Group IPT after stratification by sex: 134 women (82.7%) and 28
men (17.3%). Both treatments were manual based and consisted of
20 weekly 90-min group sessions and of 3 individual sessions.
Groups included 9 patients each and were led by PhD therapists
and by cotherapists who were at least advanced doctoral students.
All patients signed an informed consent approved by the site-
respective institutional review board.
Of the 162 randomized patients, 146 (90.1%) completed treat-
ment and 16 (9.9%) dropped out before the end of treatment.
Analyses were based on assessment completers: Of the random-
ized patients, 158 (97.5%) completed posttreatment assessment
and 143 (88.3%) completed assessment at 1-year follow-up.
Assessments
The main outcome criterion of nonresponse to treatment was
operationalized as nonabstinence, which was defined as having
one or more episodes of binge eating (i.e., eating an unusually
large amount of food, accompanied by a sense of loss of control;
American Psychiatric Association, 2000) in the past 28 days (Eat-
ing Disorder Examination [EDE] 12.0D; Fairburn & Cooper,
1993).
Predictor variables were derived from a structured interview and
from self-report questionnaires. Potential pretreatment predictors,
which were assessed prior to treatment initiation, included the
following: (a) duration since first onset of binge eating; (b) the
temporal order of binge eating versus dieting onset; (c) eating
disorder psychopathology, specifically, shape/weight concern and
restraint (EDE); (d) general psychiatric symptomatology (Global
Severity Index [GSI] from the Symptom Checklist–90 –Revised
[SCL–90 –R]; Derogatis, 1977); (e) comorbid psychiatric diagno-
sis (Structured Clinical Interview for DSM–III–R [SCID, SCID II];
Spitzer, Williams, Gibbon, & First, 1990, 1992); (f) self-esteem
(Rosenberg Self-Esteem Scale [RSES]; Rosenberg, 1979); (g)
negative affectivity subtype (cluster analytically derived from
EDE restraint, GSI, and Rosenberg Self-Esteem Scale; Stice et al.,
2001); (h) interpersonal problems (Inventory of Interpersonal
Problems [IIP]; Horowitz, Rosenberg, Baer, Ureno, & Villasenor,
1988); (i) social functioning (Social Adjustment Scale; Weissman
& Bothwell, 1976); (j) site (New Haven versus San Diego); (k)
assignment to and perceived suitability of CBT versus IPT; (l)
sociodemographic characteristics; and (m) body mass index (kg/
m
2
), which was calculated from measured height and weight.
Potential intreatment predictors of treatment nonresponse were
assessed immediately following Group Sessions 6 and 10 (i.e., at early
treatment and at midtreatment); they included group cohesion (Group
Attitude Scale [GAS]; Evans & Jarvis, 1986), group climate (En-
gaged, Avoiding, and Conflict Scales from the Group Climate Ques-
tionnaire; MacKenzie, 1981), and psychotherapeutic alliance (Cali-
fornia Psychotherapy Alliance Scale; Gaston, 1991). Other potential
intreatment predictors were assessed following Session 10 only (i.e.,
at midtreatment); they included eating disorder psychopathology (Eat-
ing Disorder Examination–Questionnaire [EDE–Q]; Fairburn & Beg-
lin, 1994), interpersonal problems (IIP), general psychopathology
(GSI), and self-esteem (RSES). For these latter constructs, midtreat-
ment levels and difference scores between pre- and midtreatment
were used. In addition, assignment to CBT versus IPT, attrition from
treatment (see Participants and Procedure), and number of sessions
attended were used as predictors in order to capture treatment speci-
ficity and dose–response relationship.
Data Analytic Plan
Signal detection analysis was used to identify distinct patient
subgroups likely to show treatment nonresponse based on pretreat-
ment and intreatment characteristics (Kraemer, 1992). Signal de-
tection analysis is a well-established procedure that is hypothesis
generating rather than hypothesis testing, nonparametric and
distribution-free, and allows for consideration of large sets of
predictors, while being robust to multicollinearity, outliers, and
missing data problems (Kiernan, Kraemer, Winkleby, King, &
Taylor, 2001). Sensitive to detecting interactions between predic-
tors, it is well-suited to clinical decision making, as algorithms are
derived for identification of patients at risk of treatment nonre-
sponse (Agras et al., 2000). The signal detection analytic method
of calculating receiver operating characteristics (ROC) and recur-
sive partitioning was applied; all predictor variables were included,
regardless of their zero-order associations. First, for each variable,
cutoff points were determined to split the sample into likely
treatment nonresponse versus response by computing sensitivity
and specificity. Next, equally weighting those cutoff points for all
variables, the optimal predictor and cutoff point was identified,
creating sample subsets with a predicted negative versus positive
outcome; this process was repeated on the identified sample sub-
sets, and so on, using chi-square tests at p ⬍ .01 as a stopping rule
646
BRIEF REPORTS
or proceeding until there were n ⬍ 10 individuals by subset. Four
ROC analyses were conducted, including two outcomes (nonre-
sponse at posttreatment and at 1-year follow-up) and two sets of
predictors (pre- and intreatment variables). Pre- and intreatment
predictors were analyzed separately in order to facilitate clinical
decision-making prior to treatment or during the early treatment
phase. The ROC-derived subgroups were further characterized on
all pretreatment and all intreatment variables (analyses of variance
and Tukey honestly significant difference tests or chi-square tests,
respectively; p ⬍ .01). Effect size of ROC-derived classification
was evaluated with the area under the curve (AUC) statistic
(Kraemer & Kupfer, 2006).
1
Results
ROC Analyses: Nonresponse at Posttreatment
Results from ROC analyses are depicted in Figure 1. Of 158
patients, 35 (22.1%) showed posttreatment nonresponse. ROC anal-
ysis for pretreatment predictors revealed that posttreatment nonre-
sponse was best predicted by an initial IIP score ⱖ 1.7 (indicative of
greater interpersonal problems) or, in the case of a lower IIP score, a
combined shape/weight concern score on the EDE ⱖ 4.5 (indicative
of more severe shape/weight concern; both ps ⬍ .01), overall test,
2
(1, N ⫽ 156) ⫽ 15.29, p ⬍ .001. For prediction of posttreatment
nonresponse by intreatment variables, ROC analysis found nonre-
sponse best predicted by a midtreatment score on the IIP ⱖ 1.9,
among individuals with lower IIP scores, a midtreatment EDE shape/
weight concern score ⱖ 3.1 (both ps ⬍ .01), overall test,
2
(1, N ⫽
149) ⫽ 18.80, p ⬍ .001.
ROC Analyses: Nonresponse at 1-Year Follow-Up
As depicted in Figure 1, 43 out of 143 patients (30.1%) showed
nonresponse at 1-year follow-up. ROC analysis of pretreatment
characteristics identified a pretreatment IIP score ⱖ 1.3 as a
significant predictor of follow-up nonresponse ( p ⬍ .01). ROC
analysis of intreatment characteristics revealed that follow-up non-
response was best predicted by a GAS score ⬍ 148 at Session 6
(indicative of lower perceived group cohesion; p ⬍ .01).
Clinical Utility of Algorithms
The algorithms derived from ROC analyses had adequate sensitiv-
ity and specificity (see Figure 1). If used to identify patients at risk of
nonresponse to the standard treatments of CBT or IPT, and to thus
select them for differential or augmented care, 64.8%–73.2% would
correctly be assigned to standard, to differential, or to augmented care
(true positives, true negatives), 18.1%–23.2% would unnecessarily be
assigned to differential or to augmented care (false positives), and
7.4%–15.2% would be assigned to standard treatment but not respond
(false negatives). Effect size of algorithm-based classification was
mostly medium (61.8% ⱕ AUC ⱕ 65.5%).
Clinical Profiles of Subgroups Identified by ROC Analyses
As presented in the left side of Table 1, patients with high inter-
personal problems, who were likely to show posttreatment nonre-
sponse, had greater general psychopathology and lower self-esteem at
pre- and midtreatment, higher pretreatment rates of any personality
disorder, of any Cluster B personality disorder, and of a high negative
affectivity subtype, and they also displayed lower initial social adjust-
ment, compared with patients who had low interpersonal problems
and low shape/weight concerns (all ps ⬍ .01). Those patients who had
low interpersonal problems but high shape/weight concerns had an
intermediate position between the other two groups; their level of
shape/weight concerns persisted through midtreatment, whereas it
decreased from pre- to midtreatment in both other groups ( p ⬍ .01).
Patients with high interpersonal problems, who were likely to show
nonresponse at 1-year follow-up, had a similar constellation of mood-
and personality-related symptoms that coincided with low perceived
cohesion of the therapeutic group and with low engagement in the
therapeutic group (see Table 1, right side; all ps ⬍ .01).
Discussion
The current study examined pre- and intreatment characteristics as
predictors of poor treatment outcome in a large, randomized trial of group
CBT and group IPT for individuals with BED. Using ROC analyses,
greater extent of interpersonal problems prior to treatment initiation or at
midtreatment emerged as a major negative prognostic indicator, predict-
ing posttreatment and long-term nonresponse. Greater shape and weight
concerns, among those with low interpersonal problems, were also pre-
dictive of posttreatment nonresponse. In addition, lower group cohesion
in the early treatment phase emerged as a process-related predictor of
long-term nonresponse.
A higher level of interpersonal problems and less perceived group
cohesion were thus both central in determining group treatment out-
come and were likely intertwined. Patients with a higher level of
interpersonal dysfunction and with a greater level of related general
psychopathology and personality disturbance also perceived less
group cohesion. The latter finding parallels the predictive value of
early therapeutic alliance in individual treatment of bulimia nervosa
(Loeb et al., 2005; Wilson et al., 1999). The present results further
indicate that interpersonal problems and shape and weight concerns
may be differentially important for the maintenance of binge eating in
subgroups of patients with BED (Fairburn, Cooper, & Shafran, 2003),
which points to the necessity of further understanding and delineating
the heterogeneity associated with this disorder. Although patients with
higher interpersonal problems or higher shape and weight concerns
had a greater likelihood of poor treatment outcome, it should be noted
that overall, CBT and IPT significantly improved these and other
psychological symptoms associated with binge eating abstinence, the
most rigorous outcome criterion (Wilfley et al., 2002).
The predictors of treatment nonresponse are valid for both treat-
ments; treatment-specific moderators or mediators were not identi-
1
The AUC statistic estimates the probability that a randomly selected
patient with a positive test (e.g., a GAS score ⬍ 148 at Session 6 for the
prediction of treatment nonresponse at 1-year follow-up) will more likely show
nonresponse to treatment than will a patient with a negative test (i.e., GAS ⱖ
148). The AUC statistic can similarly be interpreted for continuous and for
categorical variables, and effect-size classification reflects that of Cohen’s d
(low, AUC ⬍ 63.8%, Cohen’s d ⬍ 0.5; medium, 63.8% ⱕ AUC ⬍ 71.4%,
0.5 ⱕ Cohen’s d ⬍ 0.8; large, AUC ⱖ 71.4%, Cohen’s d ⱖ 0.8). An AUC ⫽
50.0% indicates that a patient with a positive test is just as likely to show
treatment nonresponse as is a patient with a negative test, whereas an AUC ⫽
100.0% indicates that every patient with a positive test shows treatment
nonresponse and that every patient with a negative test shows treatment
response (for further details, see Kraemer & Kupfer, 2006).
647
BRIEF REPORTS
fied.
2
As both treatments
were equally intense, were delivered in
group format, were adapted to BED, and were equally potent, sub-
groups of patients who responded differently to CBT versus IPT may
not have emerged. Treatment-specific mechanisms of action may not
have been identified, because change in binge eating or in the asso-
ciated psychopathology earlier than at midtreatment was not assessed.
In fact, as suggested by examinations of mediators and of time course
in comparative treatment trials of bulimia nervosa and of BED,
published after the current study was designed (Grilo et al., 2006;
Wilson et al., 1999, 2002), treatment specificity may more likely be
found in the course of binge eating and of associated psychological
symptoms during the initial treatment sessions, for example, through
the 1st month of treatment.
Specific algorithms for identifying patients who require addi-
tional clinical attention were derived. These algorithms could be
used in potential targeted selection of patients into differential or
augmented care, which would improve patient response to psy-
chological group therapy for BED. Utility of algorithms was
substantial, leading to correct selections of more than two thirds of
2
In a randomized clinical trial, a moderator of treatment is a pretreat
-
ment variable that is uncorrelated with treatment and that has an interactive
effect with treatment condition for predicting intervention response; thus,
a moderator of treatment indicates for whom or under what conditions a
treatment works. A mediator of treatment is a process variable that is
correlated with treatment and that has a main or interactive effect with it on
outcome; thus, a mediator of treatment indicates why and how a treatment
works. A predictor is defined here as a variable that precedes and has a
main effect on outcome but that has no interactive effect with treatment
(see Kraemer, Wilson, Fairburn, & Agras, 2002).
N
= 143
30.1% Nonresponders (43/143)
IIP
≥ 1.3, N = 59
44.1% Nonresponders (26/59)
χ
2
(1, N = 142) = 9.09, p < .01
IIP < 1.3, N = 83
20.5% Nonresponders (17/83)
High Interpersonal ProblemsLow Interpersonal Problems
N
= 158
22.1% Nonresponders (35/158)
Shape/Weight Concerns ≥ 3.1
46.8% Nonresponders (7/19)
Shape/Weight Concerns < 3.1
11.1% Nonresponders (11/99)
IIP
≥
1.9,
N = 31
45.2% Nonresponders (14/31)
χ
2
(1, N = 149) = 13.02,
p
< .001
IIP < 1.9, N = 118
15.3% Nonresponders (18/118)
χ
2
(1, N = 118) = 8.16, p
< .01
High Interpersonal Problems
Low Interpersonal Problems,
Low Shape/Weight Concerns
Low Interpersonal Problems,
High Shape/Weight Concerns
N
= 143
30.1% Nonresponders (43/143)
χ
2
(1, N = 138) = 6.94, p < .01
GAS
≥ 148,
N
= 93
22.6% Nonresponders (21/93)
GAS < 148, N
= 45
44.4% Nonresponders (20/45)
High Group Coherence
Low Group Coherence
(a) Pretreatment Predictors of Posttreatment Non-Response (b) Pretreatment Predictors of Non-Response at One-Year Follow-Up
(c) Intreatment Predictors of Posttreatment Non-Response (d) Intreatment Predictors of Non-Response at One-Year Follow-Up
N = 158
22.1% Nonresponders (35/158)
Shape/Weight Concerns ≥ 4.5
40.0% Nonresponders (6/15)
Shape/Weight Concerns < 4.5
13.3% Nonresponders (14/105)
IIP ≥
1.7,
N = 36
41.7% Nonresponders (15/36)
Interpersonal Problems (IIP)
χ
2
(1, N = 156) = 9.95, p
< .01
IIP < 1.7, N = 120
16.7% Nonresponders (20/120)
χ
2
(1, N = 120) = 6.72, p < .01
High Interpersonal Problems
Low Interpersonal Problems,
Low Shape/Weight Concerns
Low Interpersonal Problems,
High Shape/Weight Concerns
Interpersonal Problems (IIP, pretreatment)
Interpersonal Problems (IIP, pretreatment)
Interpersonal Problems (IIP, midtreatment)
Group Coherence (GAS, session 6)
Shape/Weight Concerns (EDE, midtx)
Shape/Weight Concerns (EDE, pretx)
Figure 1. ROC analyses: Prediction of posttreatment and long-term nonresponse from pretreatment and intreatment
characteristics in group cognitive– behavioral therapy and in group interpersonal psychotherapy of binge eating
disorder. ROC ⫽ receiver operating characteristics; IIP ⫽ Inventory of Interpersonal Problems (range: 0–4*; scores
indicating less favorable conditions are asterisked); EDE ⫽ Eating Disorder Examination (range; 0 – 6*); GAS ⫽
Group Attitude Scale (sum score range: 20*–180); AUC ⫽ area under the curve as a measure of effect size (low:
AUC% ⬍ 63.8%, medium: 63.8% ⱕ AUC% ⬍ 71.4%, large: AUC% ⱖ 71.4%). (a) Sensitivity: 0.60; specificity:
0.75; false positives: 19.2%; false negatives: 9.0%; AUC: 64.0%; 2 missing values. (b) Sensitivity: 0.66;
specificity: 0.75; false positives: 19.5%; false negatives: 7.4%; AUC: 65.5%; 1 missing value. (c) Sensitivity:
0.60; specificity: 0.67; false positives: 23.2%; false negatives: 12.0%; AUC: 61.8%; 9 missing values. (d)
Sensitivity: 0.49; specificity: 0.74; false positives: 18.1%; false negatives: 15.2%; AUC: 65.5%; 5 missing
values.
648
BRIEF REPORTS
Table 1
Outcome, Predictor, and Profile Variables for ROC Subgroups in the Prediction of Posttreatment and Long-Term Nonresponse to
Group Cognitive Behavioral Therapy (CBT) and Group Interpersonal Psychotherapy (IPT) for Binge Eating Disorder
(a) Pretreatment prediction of posttreatment nonresponse
(b) Pretreatment prediction of
nonresponse at 1-year follow-up
Interpersonal problems Interpersonal problems
Low High Low High
Shape concern
Low High
N ⫽ 105, 67.3% N ⫽ 15, 9.6% N ⫽ 36, 23.1% N ⫽ 83, 58.5% N ⫽ 59, 41.5%
Outcome variable
Nonabstinence from binge eating, n (%) 14
a
(13.3)
6
a,b
(40.0)
15
b
(41.7)
17
a
(20.5)
26
b
(44.1)
Pretreatment predictor variables, M (SD)
IIP 0.9
a
(0.4)
1.0
a
(0.3)
2.1
b
(0.3)
0.8
a
(0.3)
1.8
b
(0.4)
Shape/weight concern composite EDE 3.2
a
(0.9)
4.9
b
(0.4)
4.0
c
(0.8)
3.5 (1.0) 3.6 (0.9)
Pretreatment profile variables
Treatment preference, CBT, n (%) 51 (49.5) 9 (60.0) 11 (31.4) 48
a
(58.5)
19
b
(32.8)
Current major depression (SCID I), n (%) 14 (13.3) 2 (13.3) 9 (25.0) 7
a
(8.4)
15
b
(25.4)
Any personality disorder (SCID II), n (%) 22
a
(21.0)
6
a
(40.0)
30
b
(83.3)
20
a
(24.1)
33
b
(55.9)
Any Cluster B personality disorder (SCID
II),
n (%) 9
a
(8.6)
1
a,b
(6.7)
10
b
(27.8)
6 (7.2) 11 (18.6)
High negative affect subtype
d
, n (%)
11
a
(10.5)
5
a,b
(33.3)
29
b
(80.6)
6
a
(7.2)
34
b
(57.6)
Global Severity Index T score (SCL-90-R),
M (SD) 39.4
a
(6.3)
43.1
a
(5.4)
52.2
b
(7.5)
38.4
a
(6.3)
48.3
b
(7.3)
RSES, M (SD) 28.9
a
(5.2)
26.5
a,b
(5.0)
22.2
b
(4.3)
29.4
a
(5.0)
23.8
b
(4.8)
SAS, M (SD) 2.0
a
(0.5)
2.3
a,b
(0.6)
2.4
b
(0.5)
2.0
a
(0.5)
2.3
b
(0.5)
(c) Intreatment prediction of posttreatment nonresponse
(d) Intreatment prediction of
nonresponse at 1-year follow-up
Interpersonal problems Group cohesion
Low High High Low
Shape concern
Low High
N ⫽ 99, 66.4% N ⫽ 19, 12.8% N ⫽ 31, 20.8% N ⫽ 93, 67.4% N ⫽ 45, 32.6%
Outcome variable, n (%)
Nonabstinence from binge eating (EDE) 11
a
(11.1)
7
b
(36.8)
14
b
(45.2)
21
a
(22.6)
20
b
(44.4)
Intreatment predictor variables, M (SD)
IIP 1.1
a
(0.4)
1.2
a
(0.4)
2.2
b
(0.3)
1.2
a
(0.6)
1.6
b
(0.6)
Change shape/weight concern composite
(EDE–Q)
f
1.7
a
(0.7)
3.5
b
(0.5)
2.1
a
(0.8)
2.0 (1.0) 2.0 (0.8)
GAS 152.2 (21.2) 156.9 (16.2) 142.3 (28.6) 164.3
a
(8.5)
124.4
b
(15.5)
Intreatment profile variables, M (SD)
Early treatment, Session 6
Group Climate Engaged (GCQ) 5.5 (0.8) 5.8 (0.7) 5.4 (0.8) 5.8
a
(0.6)
5.2
b
(0.8)
Midtreatment, Session 10
GAS 151.4 (25.2) 149.9 (24.9) 146.7 (26.4) 158.3
a
(21.4)
137.7
b
(25.2)
Global Severity Index t score (SCL-90-R) 35.7
a
(8.5)
33.3
a
(8.4)
50.1
b
(7.2)
36.4
a
(10.3)
41.2
b
(8.6)
RSES 29.9
a
(5.3)
27.7
a
(5.4)
22.7
b
(4.6)
28.9 (5.6) 27.2 (5.1)
Pre- to midtreatment change
e
Change shape/weight concern composite
(EDE)
1.3
a
(1.0)
⫺0.1
b
(0.8)
1.0
a
(1.1)
1.1 (1.1) 1.1 (1.2)
Note. The outcome variable nonabstinence from binge eating over the past 28 days was assessed for (a) and (c) at posttreatment and for (b) and (d) at 1-year
follow-up. Results for predictors identified by ROC analysis are in bold type. Only profile variables indicating group differences significant at p ⬍ .01 are presented
(analyses of variance and Tukey honestly significant difference tests or chi-square tests, respectively). Missing values: (a), 2; (b), 1; (c), 9; (d), 5. IIP ⫽ Inventory
of Interpersonal Problems (range: 0 – 4
*
; scores indicating less favorable conditions are asterisked); EDE ⫽ Eating Disorder Examination (0 – 6
*
); SCID I, II ⫽
Structured Clinical Interview for DSM-III-R; SCL-90-R ⫽ Symptom Checklist-90-Revised (Global Severity Index T score ⱖ 63 as cutoff score for clinically
significant psychopathology); RSES ⫽ Rosenberg Self-Esteem Scale (sum score range: 10
*
– 40); SAS ⫽ Social Adjustment Scale (1–5
*
); GAS ⫽ Group Attitude
Scale (sum score range: 20
*
–180); GCQ ⫽ Group Climate Questionnaire (1
*
–7); EDE–Q ⫽ Eating Disorder Examination–Questionnaire (0 – 6
*
).
a,b,c
Different superscripts indicate significant group differences ( p ⬍ .01).
d
Negative affectivity subtyping was cluster analytically derived from EDE
restraint, SCL-90-R Global Severity Index, and RSES.
e
Change scores between pre- and midtreatment were calculated as pretreatment score minus
midtreatment score.
f
The EDE was administered at pre- and at midtreatment.
649
BRIEF REPORTS
patients and to selection for unnecessary augmented care in less
than one fourth of patients (only ⱕ 15.2% false negatives). Using
ROC analyses on random splits of the current sample, the same
predictors, or correlates of them, were confirmed as negative
prognostic indicators.
3
Nevertheless, validation of derived algo-
rithms in an independent sample is needed.
Clinically, our results indicate that patients whose interpersonal
dysfunction is similar to or greater than population norms for
psychiatric disorders (e.g., those with personality disturbances;
Horowitz et al., 1988; Wilfley et al., 2000) and patients with low
interpersonal dysfunction who suffer from severe shape and
weight concerns may need differential or augmented care. Special
interventions on interpersonal disturbance (e.g., Markowitz,
Skodol, & Bleiberg, 2006) and on body image disturbance (e.g.,
Fairburn et al., 2003) should be considered for these patient
subgroups; optimal timing and sequencing of such targeted treat-
ment awaits examination. To prevent potential low perceived
group cohesion, therapists need to focus on patients’ group en-
gagement, for example, by enhancing patients’ mutual understand-
ing and by fostering positive group treatment expectations (see
Constantino, Arnow, Blasey, & Agras, 2005). Fine-grained exam-
ination of the time course of treatment, which would allow iden-
tification of moderators and of mediators, could enhance specifi-
cation of treatment components and evaluation of models of
adjunctive, extended, or sequential care (National Institute of
Clinical Excellence, 2004). Such clinical research is warranted and
would allow researchers to pinpoint the optimal treatment delivery
for likely nonresponders, key to improving these patients’ response
to psychotherapy for BED.
References
Agras, W. S., Crow, S. J., Halmi, K. A., Mitchell, J. E., Wilson, G. T., &
Kraemer, H. C. (2000). Outcome predictors for the cognitive behavior
treatment of bulimia nervosa: Data from a multisite study. American
Journal of Psychiatry, 157, 1302–1308.
Agras, W. S., Telch, C. F., Arnow, B., Eldredge, K., Detzer, M. J.,
Henderson, J., & Mornell, M. (1995). Does interpersonal therapy help
patients with binge eating disorder who fail to respond to cognitive–
behavioral therapy? Journal of Consulting and Clinical Psychology, 63,
356 –360.
Agras, W. S., Telch, C. F., Arnow, B., Eldredge, K., & Marnell, M. (1997).
One-year follow-up of cognitive– behavioral therapy for obese individ-
uals with binge eating disorder. Journal of Consulting and Clinical
Psychology, 65, 343–347.
American Psychiatric Association. (2000). Diagnostic and statistical man-
ual of mental disorders (4th ed., text rev.). Washington, DC: Author.
Carter, J. C., & Fairburn, C. G. (1998). Cognitive– behavioral self-help for
binge eating disorder: A controlled effectiveness study. Journal of
Consulting and Clinical Psychology, 66, 616–623.
Constantino, M. J., Arnow, B. A., Blasey, C., & Agras, W. S. (2005). The
association between patient characteristics and the therapeutic alliance in
cognitive– behavioral and interpersonal therapy for bulimia nervosa.
Journal of Consulting and Clinical Psychology, 73, 203–211.
Derogatis, L. R. (1977). The SCL–90 manual: I. Scoring, administration,
and procedures for the SCL–90. Baltimore: Clinical Psychometric Re-
search.
Evans, N. J., & Jarvis, P. A. (1986). The Group Attitude Scale. Small
Group Behavior, 17, 203–216.
Fairburn, C. G., Agras, W. S., Walsh, B. T., Wilson, G. T., & Stice, E.
(2004). Prediction of outcome in bulimia nervosa by early change in
treatment. American Journal of Psychiatry, 161, 2322–2324.
Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders:
Interview or self-report questionnaire? International Journal of Eating
Disorders, 16, 363–370.
Fairburn, C. G., & Cooper, P. J. (1993). The Eating Disorder Examination
(12th edition). In C. G. Fairburn & G. T. Wilson (Eds.), Binge eating:
Nature, assessment, and treatment (pp. 317–360). New York: Guilford
Press.
Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive behaviour
therapy for eating disorders: A “transdiagnostic” theory and treatment.
Behaviour Research and Therapy, 41, 509–528.
Gaston, L. (1991). Reliability and criterion-related validity of the Califor-
nia Psychotherapy Alliance Scales, Patient Version. Psychological As-
sessment, 3, 368 –374.
Grilo, C. M., Masheb, R. M., & Wilson, G. T. (2006). Rapid response to
treatment for binge eating disorder. Journal of Consulting and Clinical
Psychology, 74, 602– 613.
Horowitz, L. M., Rosenberg, S. E., Baer, B. A., Ureno, G., & Villasenor,
V. S. (1988). Inventory of interpersonal problems: Psychometric prop-
erties and clinical applications. Journal of Consulting and Clinical
Psychology, 56, 885– 892.
Kiernan, M., Kraemer, H. C., Winkleby, M. A., King, A. C., & Taylor,
C. B. (2001). Do logistic regression and signal detection identify differ-
ent subgroups at risk? Implications for the design of tailored interven-
tions. Psychological Methods, 6, 35– 48.
Kraemer, H. C. (1992). Evaluating medical tests: Objective and quantita-
tive guidelines. Newbury Park, CA: Sage.
Kraemer, H. C., & Kupfer, D. J. (2006). Size of treatment effects and their
importance to clinical research and practice. Biological Psychiatry, 59,
990 –996.
Kraemer, H. C., Wilson, G. T., Fairburn, C. G., & Agras, W. S. (2002).
Mediators and moderators of treatment effects in randomized clinical
trials. Archives of General Psychiatry, 59, 877– 883.
Loeb, K. L., Wilson, G. T., Gilbert, J. S., & Labouvie, E. (2000). Guided
3
For determining reproducibility of the identified predictors, we ran
post hoc ROC analyses using samples with two random selections of 75%
of patients in each of the four ROC analyses; thus, eight post hoc ROC
analyses on eight randomly selected samples were conducted. These split-
sample analyses identified either the same predictors (IIP, GAS), with
cutoff points similar to those of the full-sample analyses, or predictors that
were significantly correlated with them ( p ⬍ .001):
(a) Prediction of posttreatment nonresponse from pretreatment
predictor variables. Sample 1: SCL–90–R GSI ⱖ 53;
2
(1, N ⫽
117) ⫽ 15.04, p ⬍ .001; Pearson correlation coefficient with
pretreatment IIP score, r ⫽ .76, p ⬍ .001. Sample 2: no solution
( p ⬎ .01). (b) Prediction of nonresponse at 1-year follow-up from
pretreatment predictor variables. Sample 3: SCL–90 –R GSI ⱖ 45;
2
(1, N ⫽ 107) ⫽ 9.93, p ⬍ .01; Pearson correlation coefficient
with pretreatment IIP score, see above ( p ⬍.001). Sample 4: no
solution ( p ⬎ .01). (c) Prediction of posttreatment nonresponse
from intreatment predictor variables. Sample 5: midtreatment
IIP ⱖ 1.8;
2
(1, N ⫽ 116) ⫽ 16.31, p ⬍ .001. Sample 6: GAS at
Session 6 ⬍ 140;
2
(1, N ⫽ 116) ⫽ 9.58, p ⬍ .01. (d) Prediction of
nonresponse at 1-year follow-up from intreatment predictor variables.
Sample 7: decrease of IIP from pre- to midtreatment ⱖ 0.25;
2
(1,
N ⫽ 94) ⫽ 9.58, p ⬍ .01. Sample 8: no solution ( p ⬎ .01).
Thus, although these split-sample analyses expectedly did not confirm
the exact results from the full-sample analyses, the results were overall
consistent with high interpersonal problems, or with correlates of them, as
main negative prognostic indicators of posttreatment and of long-term
nonresponse to group psychotherapy for BED. The finding of no statisti-
cally significant solutions for several analyses is likely attributable to a lack
of power related to the sample split.
650
BRIEF REPORTS
and unguided self-help for binge eating. Behaviour Research and Ther-
apy, 38, 259 –272.
Loeb, K. L., Wilson, G. T., Labouvie, E., Pratt, E. M., Hayaki, J., Walsh,
B. T., et al. (2005). Therapeutic alliance and treatment adherence in two
interventions for bulimia nervosa: A study of process and outcome.
Journal of Consulting and Clinical Psychology, 73, 1097–1107.
MacKenzie, K. R. (1981). Measurement of group climate. International
Journal of Group Psychotherapy, 31, 287–295.
Markowitz, J. C., Skodol, A. E., & Bleiberg, K. (2006). Interpersonal
psychotherapy for borderline personality disorder: Possible mechanisms
of change. Journal of Clinical Psychology, 62, 431– 444.
National Institute for Clinical Excellence. (2004). Eating disorders: Core
interventions in the treatment and management of anorexia nervosa,
bulimia nervosa and related eating disorders (Clinical Guideline 9).
Retrieved August 8, 2006, from http://www.nice.org.uk
Peterson, C. B., Crow, S. J., Nugent, S., Mitchell, J. E., Engbloom, S., &
Mussell, M. P. (2000). Predictors of treatment outcome for binge eating
disorder. International Journal of Eating Disorders, 28, 131–138.
Rosenberg, M. (1979). Conceiving the self. New York: Basic Books.
Safer, D. L., Lively, T. J., Telch, C. F., & Agras, W. S. (2002). Predictors
of relapse following successful dialectical behavior therapy for binge
eating disorder. International Journal of Eating Disorders, 32, 155–163.
Spitzer, R. L., Williams, J. B., Gibbon, M., & First, M. B. (1992). The
structured clinical interview for DSM–III–R (SCID): I. Archives of
General Psychiatry, 49, 624 – 629.
Spitzer, R. L., Williams, J. B. W., Gibbon, M., & First, M. B. (1990).
User’s guide for the Structured Clinical Interview for DSM–III–R:
SCID. Washington, DC: American Psychiatric Press.
Stice, E., Agras, W. S., Telch, C. F., Halmi, K. A., Mitchell, J. E., &
Wilson, T. (2001). Subtyping binge eating-disordered women along
dieting and negative affect dimensions. International Journal of Eating
Disorders, 30, 11–27.
Weissman, M. M., & Bothwell, S. (1976). Assessment of social adjust-
ment by patient self-report. Archives of General Psychiatry, 33,
1111–1115.
Wilfley, D. E., Friedman, M. A., Dounchis, J. Z., Stein, R. I., Welch,
R. R., & Ball, S. A. (2000). Comorbid psychopathology in binge
eating disorder: Relation to eating disorder severity at baseline and
following treatment. Journal of Consulting and Clinical Psychology,
68, 641– 649.
Wilfley, D. E., Welch, R. R., Stein, R. I., Spurrell, E. B., Cohen, L. R.,
Saelens, B. E., et al. (2002). A randomized comparison of group
cognitive– behavioral therapy and group interpersonal psychotherapy for
the treatment of overweight individuals with binge-eating disorder.
Archives of General Psychiatry, 59, 713–721.
Wilson, G. T., & Fairburn, C. G. (2002). Treatments for eating disorders.
In P. E. Nathan & J. M. Gorman (Eds.), A guide to treatments that work
(2nd ed., pp. 559 –592). New York: Oxford University Press.
Wilson, G. T., Fairburn, C. G., Agras, W. S., Walsh, B. T., & Kraemer, H.
(2002). Cognitive–behavioral therapy for bulimia nervosa: Time course
and mechanisms of change. Journal of Consulting and Clinical Psychol-
ogy, 70, 267–274.
Wilson, G. T., Loeb, K. L., Walsh, B. T., Labouvie, E., Petkova, E., Liu,
X., & Waternaux, C. (1999). Psychological versus pharmacological
treatments of bulimia nervosa: Predictors and processes of change.
Journal of Consulting and Clinical Psychology, 67, 451– 459.
Wonderlich, S. A., de Zwaan, M., Mitchell, J. E., Peterson, C., & Crow, S.
(2003). Psychological and dietary treatments of binge eating disorder:
Conceptual implications. International Journal of Eating Disorders,
34(Suppl.), 58–73.
Received September 29, 2006
Revision received April 30, 2007
Accepted May 10, 2007 䡲
651
BRIEF REPORTS
