Psychotherapy Research

ISSN: 1050-3307 (Print) 1468-4381 (Online) Journal homepage: http://www.tandfonline.com/loi/tpsr20

Patient baseline interpersonal problems as

moderators of outcome in two psychotherapies
for bulimia nervosa

Juan Martin Gomez Penedo, Michael J. Constantino, Alice E. Coyne,

Samantha L. Bernecker & Lotte Smith-Hansen

To cite this article: Juan Martin Gomez Penedo, Michael J. Constantino, Alice E. Coyne,
Samantha L. Bernecker & Lotte Smith-Hansen (2018): Patient baseline interpersonal problems as
moderators of outcome in two psychotherapies for bulimia nervosa, Psychotherapy Research, DOI:
10.1080/10503307.2018.1425931

To link to this article: https://doi.org/10.1080/10503307.2018.1425931

Published online: 19 Jan 2018.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tpsr20
Psychotherapy Research, 2018
https://doi.org/10.1080/10503307.2018.1425931

EMPIRICAL PAPER

Patient baseline interpersonal problems as moderators of outcome in

two psychotherapies for bulimia nervosa

JUAN MARTIN GOMEZ PENEDO1, MICHAEL J. CONSTANTINO2, ALICE E. COYNE2,

SAMANTHA L. BERNECKER 2∗ , & LOTTE SMITH-HANSEN3
1
CONICET & Department of Psychology, Universidad de Buenos Aires, Buenos Aires, Argentina; 2Department of
Psychological and Brain Sciences, University of Massachusetts Amherst, Amherst, MA, USA & 3Department of Psychology,
Suffolk University, Boston, MA, USA
(Received 21 July 2017; revised 2 January 2018; accepted 3 January 2018)

Abstract
Objective: We tested an aptitude by treatment interaction; namely, whether patients’ baseline interpersonal problems
moderated the comparative efficacy of cognitive-behavioral therapy (CBT) vs. interpersonal psychotherapy (IPT) for
bulimia nervosa (BN). Method: Data derived from a randomized-controlled trial. Patients reported on their interpersonal
problems at baseline; purge frequency at baseline, midtreatment, and posttreatment; and global eating disorder severity at
baseline and posttreatment. We estimated the rate of change in purge frequency across therapy, and the likelihood of
attaining clinically meaningful improvement (recovery) in global eating disorder severity by posttreatment. We then tested
the interpersonal problem by treatment interactions as predictors of both outcomes. Results: Patients with more baseline
overly communal/friendly problems showed steeper reduction in likelihood of purging when treated with CBT vs. IPT.
Patients with more problems of being under communal/cold had similar reductions in likelihood of purging across both
treatments. Patients with more baseline problems of being overly agentic were more likely to recover when treated with
IPT vs. CBT, whereas patients with more problems of being under agentic were more likely to recover when treated with
CBT vs. IPT. Conclusions: Interpersonal problems related to communion and agency may inform treatment fit among
two empirically supported therapies for BN.

Keywords: interpersonal problems; moderators; cognitive-behavioral therapy (CBT); interpersonal psychotherapy (IPT);
bulimia nervosa

Clinical or methodological significance of this article: Patients with bulimia nervosa (BN) who presented with more
problems of an overly communal/friendly and/or under agentic/submissive nature showed greater improvement in
cognitive-behavioral therapy (CBT) vs. interpersonal psychotherapy (IPT). Patients with more problems of being overly
agentic/domineering were more likely to recover when treated with IPT vs. CBT, whereas patients with more problems of
being under communal/cold had a similar outcome in the two treatments. These aptitudes (trait) by treatment interactions
suggest that it is important to assess baseline interpersonal problems to help establish which of these two empirically
supported therapies would, on average, be a better fit for patients with BN.
Both cognitive-behavioral therapy (CBT) and inter- Brenner et al., 2016). It is possible that such residual
personal psychotherapy (IPT) have demonstrated symptoms, or lack of clinically significant response
general efficacy in treating bulimia nervosa (BN; altogether, could stem, at least partially, from patients
Hay & Claudino, 2010; Wilson & Shafran, 2005). being poorly suited for CBT or IPT—two vastly
However, a substantial proportion of bulimic patients different approaches for treating BN.
receiving these treatments continue to experience On the one hand, CBT focuses explicitly on BN
symptoms after therapy ends (Agras, Walsh, Fair- symptoms. The model assumes that dysfunctional
burn, Wilson, & Kraemer, 2000; Thompson- (rigid and unrealistic) beliefs about food, eating

Correspondence concerning this article should be addressed to Juan Martin Gomez Penedo CONICET & Department of Psychology, Uni-
versidad de Buenos Aires, 2353 Lavalle St., Ciudad de Buenos Aires, Argentina. Email: [email protected]
∗
Present address: Department of Psychology, Harvard University, Cambridge, MA, USA.

© 2018 Society for Psychotherapy Research

2 J. M. Gomez Penedo et al.

behaviors, and perceptions of body weight/shape problems than non-clinical control participants;
underlie the development and maintenance of the however, supporting the notion that interpersonal
disorder (Spangler, 1999). Consequently, the CBT problems in BN individuals are heterogeneous, one
clinician uses multiple directive strategies to increase study found problems to be primarily of the overly
adaptive eating, and to challenge distorted thought submissive type (Hartmann, Zeeck, & Barrett,
processes about food, weight, and shape. On the 2010), whereas the others (e.g., Hopwood et al.,
other hand, IPT centers not on BN symptoms expli- 2007) found problem profiles evenly distributed
citly, but rather on interpersonal difficulties that across the IIP-C vectors. In a study examining the
implicitly connect to the manifest pathology (Apple, association between interpersonal problems and
1999). The model assumes that interpersonal pro- bulimic episodes in a non-clinical sample, the
blems, such as role disputes or unresolved grief, pre- researchers found an indirect effect; higher levels of
cipitate BN problems. Thus, the IPT clinician uses overall interpersonal problems were related to
multiple active, yet nondirective, strategies to greater negative affect, which, in turn, was related
implement adaptive interpersonal change, increase to a higher probability of having an objective
social support, and reduce relational distress, all of bulimic episode (Ansell et al., 2012). Additionally,
which are purported to facilitate long-term several specific types of interpersonal problems
reductions in disordered eating symptoms, without related to bulimic episodes either directly or
ever discussing such symptoms directly. indirectly. First, individuals with more problems
Given that CBT and IPT represent two clearly dis- related to being overly friendly/nurturant (i.e., one
tinct approaches to treating BN, it seems important extreme end of the communion dimension) were
to identify factors that might optimize bulimic more likely to have bulimic episodes. Second,
patients’ match to one of these evidence-based thera- patients with more problems related to being overly
pies. One class of factors that might influence differ- hostile/cold (i.e., the other extreme end of the com-
ential treatment response is the theoretically relevant munion dimension) reported more negative affect
individual characteristics with which BN patients that, in turn, related to a higher probability of
present; that is, aptitude by treatment interactions bulimic episodes.
(ATIs; Accurso et al., 2016). Many BN theories Based on the literature, interpersonal problems
posit that maladaptive interpersonal functioning is appear to be candidate aptitudes (or more precisely,
implicated in the genesis and maintenance of the dis- in this case, traits) that may moderate patient
order. From this pathoplasticity perspective response to CBT and IPT for BN, which, as noted,
(Hopwood, Clarke, & Perez, 2007), individuals are the current individual psychotherapies with the
with BN have evidenced a variety of elevated social most empirical support. To date, though, research
problems compared to healthy controls (Tanofsky- on these approaches has largely focused on their
Kraff & Wilfley, 2010), and several types of interper- general efficacy (Agras et al., 2000). We are unaware
sonal difficulties have been associated with greater of studies that have examined interpersonal ATIs,
severity of BN symptomatology (e.g., Ansell, Grilo, despite the promise such a focus holds for empirically
& White, 2012). responsive treatment selection (Constantino,
To more precisely characterize prototypic interper- Boswell, Bernecker, & Castonguay, 2013). To this
sonal difficulties in persons with BN, some research- end, the present study focused on this interpersonal
ers have drawn on the Inventory of Interpersonal trait × treatment question. Given that previous
Problems-Circumplex version (IIP-C; Horowitz, research has generally failed to isolate specific prototy-
Alden, Wiggins, & Pincus, 2000), a measure of inter- pic IIP-C vectors that characterize BN, and/or has
personal excesses and inhibitions distributed in circu- implicated the foundational IIP-C dimensions of
lar fashion around the two relational dimensions of agency and communion in BN symptom severity,
agency (ranging from problematic submissive to pro- we reasoned that these dimensions would be most
blematic domineering behaviors) and communion likely to inform treatment suitability. Regarding
(ranging from problematic hostile/cold to proble- agency, it may be that BN patients with an interperso-
matic friendly/nurturing behaviors; Horowitz et al., nal style of, and need for, being overly domineering
2000). Two-dimensional combinations of agency toward others may have difficulty deferring to and fol-
and communion create eight specific problem lowing the CBT therapist’s directive style centered on
vectors that reflect being too domineering, intrusive, reducing disordered eating cognitions and behaviors
self-sacrificing, overly accommodating, nonassertive, (Fairburn, Cooper, & Shafran, 2003). Rather,
socially inhibited, cold, and/or vindictive. patients with a more domineering problem style
Across three relevant IIP-C studies with clinical may be better suited for IPT given its focus on the
BN samples, individuals reported significantly more interpersonal difficulties that underlie eating pro-
overall distress (total score) from interpersonal blems. To the extent that a pattern of problematic
 Psychotherapy Research 3

dominance emerges in the interpersonal assessment, undermining IPT’s central focus that purportedly
treatment can help patients more effectively navigate drives its efficacy).
self and other needs for agency in relationships Of course, these moderator ideas require testing,
(Wilson, 1996). With such interpersonal gains may which was our aim in this study drawing on a large
come less risk for disordered eating behavior (even sample of women with BN who were treated with
in the absence of specific work on eating-relevant individual CBT or IPT in the context of a random-
cognitions and actions). ized-controlled trial (RCT; Agras et al., 2000). We
On the contrary, patients with an interpersonal hypothesized that patients both with more baseline
style of, and need for, being overly submissive to problems of being overly submissive and overly
others may find the CBT therapist’s directive, friendly would demonstrate more improvement in
symptom-focused approach suitable to their typical CBT vs. IPT, whereas patients with more baseline
way of relating. With this complementary pattern of problems of being overly domineering and overly
therapist influence and patient deference, it is poss- cold/hostile would demonstrate more improvement
ible that CBT strategies would work well for these in IPT vs. CBT. Improvement was operationalized
patients, at least with regard to improving eating as (i) a continuous measure of purge frequency over
symptoms in the short-term. However, with excessive the course of treatment and (ii) a dichotomous
submissiveness often manifesting as treatment com- measure of clinically meaningful improvement from
pliance (Muran, Segal, Samstag, & Crawford, baseline to posttreatment on a global index of
1994), it is possible that the interpersonal focus of eating disorder symptoms.
IPT might overlook this behavior as problematic.
Thus, the IPT therapist’s focus on other interperso-
nal contents (and not specific eating symptoms, as Method
is proscribed in IPT), may miss the mark on a likely
interpersonal risk factor of limited effective assertive- Data derived from the aforementioned RCT that
ness. Consequently, these patients may do less well in compared CBT and IPT for BN across two sites:
this time-limited treatment. Stanford University and Columbia University (see
Regarding communion, BN patients with an inter- Agras et al., 2000, for additional design and
personal style of, and need for, being overly cold and primary outcome details). At posttreatment, CBT
distant from others might have difficulty working with significantly outperformed IPT in terms of the pro-
a CBT therapist who tends toward an expert stance portion of patients recovered, remitted, and
and active working relationship. In other words, the meeting community norms on eating behaviors and
CBT therapist’s bid for active collaboration may attitudes; however, across three follow-up occasions
prove too threatening to the person motivated to (4, 8, and 12 months), there were no significant
keep hostile distance from others (albeit to maladap- differences between CBT and IPT. There was less
tive consequence). With such threat could come attrition in IPT (19.1%) than CBT (30.0%), with
heightened reactance or resistance, which could the difference approaching significance.
undermine adaptive therapeutic process and gain
(e.g., Beutler, Harwood, Michelson, Song, &
Participants
Holman, 2011). Rather, patients with a colder
problem style may be better suited for IPT, where Patients were 220 women (110 at each site; by design,
the therapist might be able to target the hostility/cold- no men were recruited) meeting Diagnostic and Stat-
ness (as a factor that underlies the BN) with a more istical Manual of Mental Disorders (third edition,
nondirective stance less likely to promote negative revised; DSM-III-R; American Psychiatric Associ-
process. ation, 1987) criteria for BN. Patients were excluded
On the contrary, patients with an interpersonal if they had a severe physical or psychiatric (e.g., psy-
style of being overly communal/friendly toward chotic) disorder, were diagnosed with anorexia
others may find the CBT therapist’s directive, nervosa (AN), were already in a psychosocial treat-
symptom-focused approach suitable and effective in ment, were on psychotropic medications, were preg-
that it never challenges this pattern as problematic nant, and/or had engaged in a prior adequate dose of
(rather, it may capitalize on the need for communion, CBT or IPT. Patients averaged 28.1 years of age (SD
which will likely enhance the active collaboration that = 7.2 years), and the majority were Caucasian
is presumably central to CBT efficacy). However, if (77.0%) and either single, divorced, or widowed
the IPT therapist targets the patient’s need for com- (80.4%). The sample averaged 11.4 years (SD = 7.5
munion and close connection as problematic, the years) of bingeing and 9.8 years (SD = 6.8 years) of
patient who may fail to see this behavior as proble- purging. Twenty-two percent of patients also met cri-
matic could resist this direction (thereby teria for current major depressive disorder, and 37%
4 J. M. Gomez Penedo et al.

for a current personality disorder. Across the two changes, monitor treatment gains, process feelings
treatment conditions, patients were largely equivalent about treatment termination, and plan for coping
on baseline demographic and eating symptom vari- with future interpersonal distress. IPT for BN pro-
ables; however, CBT patients had significantly scribes explicit attention to eating patterns, compen-
more purge episodes and eating concerns than IPT satory behaviors, or body shape and weight attitudes;
patients. Several significant site differences also it also does not include self-monitoring of food intake
emerged; patients treated at Stanford were older or specific behavioral instruction.
and more likely to have been diagnosed with lifetime
substance abuse or dependence, whereas Columbia Measures
patients had a longer purging duration, were less
likely to have a history of AN, had fewer concerns Interpersonal problems. To assess the modera-
about eating and shape, and reported fewer global tor variable of interpersonal problems, patients com-
symptoms. pleted the original, 127-item version of the IIP
Therapists were seven doctoral-level psychologists (Horowitz, Rosenberg, Baer, Ureño, & Villaseñor,
and one psychiatrist who each treated similar 1988). For the current study, variables were derived
numbers of patients in both CBT and IPT. Although from the 64 items that comprise the IIP-C (Horowitz
data on therapists’ primary orientation and allegiance et al., 2000). The IIP-C is a widely used instrument
upon entering the trial are unavailable, all therapists assessing interpersonal inhibitions and excesses,
were experienced in treating eating disorders, with each item in this trial rated on a 5-point scale
received extensive training in the treatment proto- ranging from 0 to 4. Higher total scores indicate
cols, were closely supervised throughout the trial, more interpersonal problems. As previously
and delivered the treatments with adequate fidelity described, the IIP-C has eight subscales that com-
and competency (see Agras et al., 2000). Loeb et al. prise a circumplex of problematic interpersonal be-
(2005) reported high levels of observer-rated thera- havior around the interpersonal dimensions of
pist adherence to both treatment manuals (on a agency and communion. The IIP-C has evidenced
scale from 1 to 7, mean adherence was 5.52 in good psychometric properties (Horowitz et al.,
CBT and 5.03 in IPT). Furthermore, in a selected 2000), and the internal consistency alpha coefficients
subsample of cases, the independent raters (blind to in the current study were .94 for the total score and
the treatment) were able to identify correctly the ranged from .71 to .89 for the subscales.
treatment being conducted (Loeb et al., 2005). As noted, we focused on two IIP-C indices that
were computed based on the formulas that Ruiz
et al. (2004) presented. We calculated an interpersonal
Treatments
communion index by using data from six subscales
Both treatments were manual-driven and involved 19 weighted based on their proximity to the horizontal
individual outpatient sessions conducted over 20 communion axis. The formula is Communion = .25
weeks. The 50-minute sessions were administered [nurturant – cold + .71 (intrusive – vindictive –
twice weekly for the first two weeks, weekly for the socially inhibited + exploitable)]. Thus, the commu-
next 12 weeks, and biweekly thereafter. Therapists nion index represents interpersonal problems
delivered CBT according to the Fairburn, Marcus, ranging from being overly cold and distant on one
and Wilson (1993) manual, which directly targets end of the axis to being overly warm and self-sacrifi-
the main symptomatic features of BN (i.e., bingeing, cing on the other. Note that domineering and nonas-
purging, and shape and weight distortions) through sertive interpersonal problems are not included in
psychoeducation, behavior monitoring, strategies to this formula because they are neutral in terms of com-
reduce dietary restraint and irregular eating, strat- munion. The theoretical range for this index is
egies for challenging distorted thoughts and assump- −19.36 (colder and more distant) to 19.36 (warmer
tions, implementation of adaptive coping responses and more self-sacrificing). We also calculated an
to binge-eating triggers, and relapse prevention. interpersonal agency index by using data from six sub-
Therapists delivered IPT according to the scales weighted based on their proximity to the verti-
Klerman, Weismann, Rounsaville, and Chevron cal agency axis. The formula is: Agency = .25
(1984) manual adapted for BN (Fairburn, 1997). [domineering – nonassertive + .71 (intrusive + vin-
The IPT therapist initially draws a connection dictive – socially inhibited – exploitable)]. Thus, the
between patients’ interpersonal difficulties (e.g., agency index represents interpersonal problems
role disputes, role transitions, interpersonal deficits, ranging from overly submissive behavior on one end
unresolved grief) and their BN symptoms. Sub- of the axis to overly domineering behavior on the
sequently, this connection is only implied, as the other. In this case, nurturant and cold interpersonal
therapist helps patients implement interpersonal problems are not included in the formula because
 Psychotherapy Research 5

they are neutral in terms of agency. The theoretical ±2 standard errors of measurement) around each
range for this index is again −19.36 (more submiss- patient’s estimated true score at baseline. A true
ive) to 19.36 (more domineering). score is estimated as the product of the patient’s base-
line deviation score (i.e., the difference between a
given patient’s actual score at baseline and the
Purge frequency. To assess the continuous
sample mean) and test–retest reliability, plus the
outcome variable, patients reported on their 1-week
baseline sample mean (Speer, 1992). However, in
purge frequency. This variable was used similarly in
the present study, we also followed Lambert and
previous studies drawing on the present trial data
Ogles (2009) suggestion to use internal consistency
(e.g., Loeb et al., 2005). This variable was not only
rather than test–retest reliability as the measure of
measured on multiple occasions (allowing for a
reliability. After calculating each patient’s true score
more reliable assessment of change than simply pre-
at baseline, we next calculated the standard error of
posttreatment), but also possesses fewer assessment
measurement (SE = 0.45), which is equal to the
liabilities than other symptoms of eating pathology
product of the sample SD at baseline (1.12) and the
given that purge recall is less ambiguous than idiosyn-
root mean square of measurement error (i.e., 1
cratic definitions of bingeing and weight/shape con-
minus the internal consistency of the measure; α
cerns (Wilson, 1993). Further, researchers have
= .84). To be classified as having reliably changed,
argued that such brief self-report assessments can
patients needed to have a posttreatment score at
be used effectively for repeated measurement in
least two SE (i.e., 0.90) below their estimated true
trials after a BN diagnosis has been established with
score at baseline.
rigorous interview methods (Loeb, Pike, Walsh, &
To also contribute to our binary variable of clini-
Wilson, 1994). In this sample, the purge frequency
cally meaningful change on the global EDE index,
presented an adequate test–retest reliability (two-
we next created an index of clinically significant
way mixed, average measure’s Intraclass Correlation
response using Jacobson and Truax’s (1991) Cutoff
Coefficient [ICC] = .92).
C criteria, as this method is considered the current
best practice (Lambert & Ogles, 2009). To calculate
Severity of global eating disorder Cutoff C, or the point at which a person’s score is
symptomatology. To assess the dichotomous closer to the normal than the clinical range, we
outcome of clinically meaningful change, we drew used normative data from a non-clinical, commu-
on Eating Disorder Examination (EDE; Cooper & nity-sample, which reported a mean global EDE
Fairburn, 1987) data. The EDE is a comprehensive index of 0.93 (SD = 0.81; Fairburn et al., 2008), as
and well-validated structured interview that measures well as the means and standard deviation of the
multiple dimensions of eating disorder symptomatol- study sample at baseline, which represents a patho-
ogy over the previous 28 days. For the present study, logical BN population with a mean EDE index of
we created a global index of eating disorder severity 3.31 (SD = 1.12). Based on this analysis, the cutoff
(our main interest for capturing clinically meaningful point for clinically significant response was 2.32
change on a broader scale of BN problems) by points on the global EDE index.
summing the scores of four EDE subscales of Combing the EN assessment of reliable change and
dietary restraint, shape concerns, weight concerns, the Cutoff C assessment of clinically significant
and eating concerns (Fairburn, Cooper, & response on the global index, a patient was classified
O’Connor, 2008) and then dividing these results by as recovered (i.e., yes on clinically meaningful change)
four. Thus, the global EDE index had a theoretical when they passed both criteria, and not recovered (i.e.,
range from 0 to 6. The four indices have shown ade- no on clinically meaningful change) when they did
quate internal consistency (α ranging from .67 to not. For descriptive purposes, we also present percen-
.78), inter-rater reliability (Spearman’s Rho coeffi- tages for the following classifications: improved (when
cients ranging from .90 to .99), and test–retest patients passed only the EN confidence interval cri-
reliability (Spearman’s Rho coefficients ranging terion in the positive direction), unchanged (when
from .71 to .76; Rizvi, Peterson, Crow, & Agras, they did not pass the EN confidence interval cri-
2000). In the current sample, the global index had terion), or deteriorated (when they passed the EN con-
an alpha coefficient of .84.1 fidence interval criterion in the negative direction).
To create our binary variable of clinically meaning-
ful improvement on the global EDE index, we first
Procedure
used the Edwards–Nunnally (EN) method to calcu-
late an index of reliable change (Speer, 1992).2 This Potential trial patients were initially phone screened for
method defines reliable change as having surpassed eligibility. Those who were not ruled out with this
the lower limit of a confidence interval (based on screen were scheduled for an in-person baseline clinical
6 J. M. Gomez Penedo et al.

assessment with a trained research assistant. After within-patient data, such that HLM mimicked an
obtaining consent, research assistants administered intent-to-treat (ITT) sample by retaining all partici-
the EDE and the Structured Clinical Interview for pants who completed at least one purge frequency
the DSM-III-R (SCID; Spitzer, Williams, Gibbon, & rating. Thus, although some patients were missing
First, 1989). If they remained eligible, participants purge frequency data at week 1 (3%), midtreatment
completed a baseline battery of self-report measures, (26%), and posttreatment (30%), we still retained
including the IIP. Patient purge frequency was 216 of the 220 trial patients (98%) in the analyses.
recorded at week 1, midtreatment (week 10), and post- Given that multiple patients were nested within
treatment. The EDE was administered again at post- therapists, we first conducted 3-level models to test
treatment. The present analyses were conducted on for possible therapist effects on the weekly rate of
completely de-identified, archival trial data. change and in posttreatment level of purge frequency.
The intraclass correlations calculated from these
models suggested that therapists accounted for <1%
Results of the variability in both outcomes. Given this lack
Sample descriptives of between-therapist variability, we fit 2-level
models for our primary analyses.
We present in Online Supplemental Table 1 the
descriptive statistics of baseline interpersonal pro-
blems, purge frequency, and EDE scores across the Unconditional model. We first fit a 2-level
total sample and by each treatment condition. unconditional model with time expressed in weeks
There were no significant differences between CBT and centered at posttreatment as the only predictor.
and IPT patients’ IIP-C total score, IIP-C agency Given that purge frequency is a count variable and
index, or IIP-C communion index. At baseline, it had a positively skewed distribution across all
across both treatments, agency and communion measurement occasions (as is typical for count vari-
had a small, but significant negative correlation, r = ables), we used a Poisson (constant exposure) distri-
−.19, p = .005, with one index explaining less than bution for the outcome variable (the equation for this
4% of the variance in the other. Despite this corre- model is described in the Online Supplement).
lation, there were no evidences of problematic colli- Results of this unconditional model indicated that
nearity between these two interpersonal indexes the average likelihood of purging at posttreatment
(Tolerance = .97, VIF = 1.04). was significantly different from 0 (γ00 = 0.33, SE =
Regarding outcome measures, there were no signifi- 0.17, p = .05). For ease of interpretation, we expo-
cant differences between the two conditions on base- nentiated the coefficient to represent a rate ratio
line global EDE score. However, IPT patients (RR), which indicated that the average posttreatment
reported greater purge frequency at baseline than likelihood of another purge episode was 1.39. Also,
CBT patients. Given this difference, we controlled the average weekly reduction in the likelihood of
for baseline differences in purge frequency, as assessed purging significantly differed from 0 (γ00 = −0.10,
with the EDE, in all analyses with purge frequency as SE = 0.008, p < .001; RR = 0.90). In this case, the
the outcome variable. Additionally, because Agras RR of 0.90 indicated that patients experienced a
et al. (2000) reported an effect of site on patient 10% reduction in purge likelihood per week, on
eating disorder symptomatology, we also included average. Random effects indicated that there was sig-
site as a control variable in all of our primary analyses. nificant between-patient variability in both the
Regarding clinically meaningful improvement percen- average likelihood of purging at posttreatment (τ0 =
tages at posttreatment, 42.7% of the sample were 3.05, p < .001) and the average weekly change in
recovered (CBT = 45.5%, IPT = 40.0%), 5.5% this likelihood (τ1 = 0.005, p < .001) that could be
improved (CBT = 5.5%, IPT = 5.5%), 50.9% explained by the addition of predictors.
unchanged (CBT = 48.2%, IPT = 53.6%), and 0.9%
deteriorated (CBT = 0.9%, IPT = 0.9%).3
Conditional models. First, for comparison pur-
poses, we ran a main effects and covariates model
Primary Outcome: Purge Frequency
with treatment condition (IPT = 0, CBT = 1), the rel-
To test the two IIP dimensions (i.e., communion and evant interpersonal problem dimension (commu-
agency) as moderators of treatment efficacy on the nion, agency), site (Stanford = −0.5, Columbia =
primary purge frequency outcome, we used hierarch- 0.5), and baseline purge frequency as the level-2 pre-
ical linear modeling (HLM; Raudenbush & Bryk, dictors. Second, we ran two separate interaction
2002), which addresses dependency in repeated models, each of which added an interpersonal
measures data and is robust to handling missing problem dimension by treatment interaction as a
 Psychotherapy Research 7

level-2 predictor (the equations for these models are centered main effects and newly generated inter-
described in the Online Supplement).4 action terms were then entered as predictors in two
Results of both the main effect and interaction additional HLM models (one for high communion
models for each interpersonal problem index are pre- and one for low communion).6 The results indicated
sented in Table I. The interactive effect of treatment that for overly communal patients, there was a signifi-
by agency did not significantly predict the posttreat- cant difference between the two treatments in
ment purge likelihood or weekly change in purge like- purging rates at posttreatment (γ01 = −1.85, p
lihood across treatment. However, the interactive < .001) and in the weekly rate of change in purging
effect of treatment by communion significantly pre- rates (γ11 = −0.10, p < .001); expressed in terms of
dicted both the likelihood of purging at posttreat- RRs, when treated with CBT, overly communal/
ment, γ04 = −0.16, SE = 0.07, CI95 [−0.29, −0.03], friendly patients’ rates of purging at posttreatment
t(210) = −2.361, p = .02, pseudo R 2 = 0.07, and were .16 times as likely (i.e., 84% less likely), and
weekly change in the likelihood of purging, γ13 = the weekly rate of change in that likelihood was
−0.01, SE = 0.003, CI95 [−0.02, −0.004], t(210) = 0.90 times steeper (i.e., a 10% greater weekly
−3.139, p = .002, pseudo R 2 = 0.11.5 Overly commu- reduction in purge likelihood) than for IPT patients.
nal patients experienced lower overall likelihood of In other words, the advantage of CBT vs. IPT was
purging at posttreatment and a steeper weekly more pronounced for over communal patients than
reduction in purge likelihood when treated with for patients with an average level of communion.
CBT vs. IPT. Under communal patients had lower For under communal/cold patients, there was no sig-
likelihood of purging at posttreatment when treated nificant difference between the two treatments in
with CBT vs. IPT, though these patients had likelihood of purging at posttreatment (γ01 = −0.65,
similar reductions in likelihood of purging across p = .08), or in the weekly change in purge likelihood
both treatments (see Figure 1). (γ11 = −0.02, p = .30). In other words, the advantage
To probe further the interactive effects, we gener- of CBT vs. IPT was reduced to non-significance for
ated conditional slopes representing the effect of under communal patients. Thus, although there
treatment on weekly purge reduction at high (1 SD was a comparative advantage of CBT vs. IPT for
above the mean) versus low levels (1 SD below the patients who had either average or high levels of com-
mean) of communion. We generated these slopes munion at baseline, the two treatments were rela-
by first re-centering the relevant IIP-C index at ± 1 tively equivalent for under communal patients.
SD from the mean and then creating new interaction
terms using each of these re-centered indices. The re-
Secondary Outcome: Clinically Meaningful
Improvement
We next used SPSS 23 to conduct separate logistic
regression models examining the interactive effect of
each interpersonal problem index by treatment con-
dition on the likelihood of achieving clinically mean-
ingful improvement on the global EDE index. To
address missing values in the global EDE index at
posttreatment (26%), we used an ITT approach in
which patients’ baseline EDE global index scores
were carried forward to replace missing data at post-
treatment. We also included site as a covariate,
because we could not residualize out the effect of site
from the clinically meaningful improvement variable
(due to its binary nature). The results of these
models are presented in Online Supplemental Table 2.
There was no significant interactive effect of treat-
Figure 1. Evolution of the likelihood of purging during treatment in ment by communion on the likelihood of having clini-
overly and under communal patients, comparing CBT vs. IPT. cally meaningful improvement, β = 0.123, SE = 0.08,
Overly and under communal were defined as ±1 SD (3.72) from Wald(1) = 2.539, p = .11. However, there was a signifi-
the mean (1.66) of the communion dimension subscale in the cant interactive effect of treatment by agency on clini-
sample. Note. OV COM = overly communal; UN COM = under cally meaningful improvement, β = −0.165, SE = 0.07,
communal; CBT = cognitive-behavioral therapy; IPT = interperso-
nal psychotherapy. ∗ Conditional slopes analysis showed a signifi- Wald(1) = 6.157, p = .01, odds ratio (OR) = .85. In
cant treatment effect in overly communal patients (p < .05), but IPT patients, a 1-unit increase in agency was associ-
not in under communal patients. ated with a 1.07 times greater odds of achieving
8 J. M. Gomez Penedo et al.
Table I. Unconditional and conditional models analyzing the main and moderating effects of interpersonal problems on the likelihood of
purging.

Unconditional
model Main effect models Interactive effect models

Fixed model effects β SE β SE β SE

Likelihood of purging at posttreatment

Communion IP
Intercept 0.33∗ 0.17 1.03∗∗∗ 0.17 0.99∗∗∗ 0.17
Treatment −1.34∗∗∗ 0.26 −1.25∗∗∗ 0.25
IP −0.03 0.03 0.04 0.04
Site −0.23 0.03 −0.30 0.25
Baseline purge frequency 0.02∗∗∗ 0.003 0.02∗∗∗ 0.003
IP × treatment −0.16∗∗ 0.07
Pseudo R 2 = .07
Agency IP
Intercept 0.33∗ 0.17 1.00∗∗∗ 0.18 1.02∗∗∗ 0.10
Treatment −1.32∗∗∗ 0.26 −1.31∗∗∗ 0.26
IP 0.009 0.03 −0.02 0.04
Site −0.25 0.26 −0.27 0.26
Baseline purge frequency 0.02∗∗∗ 0.003 0.02 0.003
IP × treatment 0.05 0.06
Pseudo R 2 = .02
Weekly change on likelihood of purging
Communion IP
Intercept −0.10∗∗∗ 0.008 −0.07∗∗∗ 0.009 −0.07∗∗∗ 0.008
Treatment −0.06∗∗∗ 0.01 −0.06∗∗∗ 0.01
IP −0.002 0.002 0.002 0.002
Site −0.004 0.01 −0.008 0.01
Baseline purge frequency 0.0004∗∗ 0.0001 0.0004∗∗ 0.0001
IP × treatment −0.01∗∗ 0.003
Pseudo R 2 = .11
Agency IP
Intercept −0.10∗∗∗ 0.008 −0.07∗∗∗ 0.009 −0.07∗∗∗ 0.009
Treatment −0.06∗∗∗ 0.01 −0.06∗∗∗ 0.01
IP 0.0009 0.001 −0.0004 0.002
Site −0.005 0.01 −0.006 0.01
Baseline purge frequency 0.0004∗∗ 0.0001 0.0004∗∗ 0.0001
IP × treatment 0.003 0.003
Pseudo R 2 = .02
Model comparison for communion IP χ (8) = 177.90, p < .001 χ 2(2) = 12.94, p = .002
2

Model comparison for agency IP χ 2(8) = 176.96, p < .001 χ 2(2) = 0.95, p > .50

Note: IP = interpersonal problems; Pseudo R 2 = effect size measure based on the variance explained by the interactive effect.
∗∗∗
p < .001.
∗∗
p < .05.
∗
p = .05.

clinically meaningful improvement, whereas in CBT potential moderators of the comparative posttreat-
patients that increase in agency was related to a 0.91 ment efficacy of CBT and IPT for BN. In the
times lower odds of clinically meaningful improve- trial from which the data derived, CBT, on
ment. In other words, overly agentic patients were average, significantly outperformed IPT at post-
more likely to achieve clinically meaningful improve- treatment on indices of recovery, remittance, and
ment in IPT vs. CBT, whereas under agentic patients meeting community norms on eating behaviors
were more likely to achieve clinically meaningful and attitudes (Agras et al., 2000). These results,
improvement in CBT vs. IPT (see Figure 2).7 though, said nothing of specific patient-treatment
fit, which prompted the present ATI analyses. As
hypothesized, there was an interactive effect of
Discussion interpersonal communion problems by treatment
We examined two dimensions of patient baseline on the purge frequency outcomes. This effect was
interpersonal problems (communion and agency) as largely driven by patients with more baseline
 Psychotherapy Research 9

it is possible that the observed pattern is meaningful

beyond statistical power issues; thus, we speculate
next on what this pattern of results might mean
clinically.
Most predominantly, we draw on the reactance
and resistance literatures (e.g., Beutler et al., 2011).
Regarding communion, as we hypothesized, for
overly communal BN patients, CBT therapists may
be less likely to promote reactance or resistance
given that they are unlikely to target these patients’
interpersonal traits directly. Without challenging
this social exchange pattern as a problem underlying
the eating symptoms, the CBT therapist may actually
capitalize on the patient’s need for communion in the
form of close and active collaboration toward chan-
ging the specific behavior of purging. Indeed, pre-
Figure 2. Likelihood of having a clinical significant change in
patients with low, average, and high agency levels, comparing vious research shows that patients who are more
CBT vs. IPT. Low and high agency was defined as ±1 SD (4.32) accommodating (even if it is maladaptive for them)
from the mean (−4.13) of agency in the sample. Likelihood was cal- are more likely to engage in collaborative, connected
culated by the formula odds ratio/(1 + odds ratio). relationships than people who are cold, inhibited,
Note: CS = clinical significant change.
and/or vindictive (Muran et al., 1994). In fact, in a
study also drawing on data from the current trial,
problems of being overly communal/friendly patients with overly communal problems reported
demonstrating lower purge rates at posttreatment better therapeutic alliances in CBT (Constantino &
and a steeper reduction in those rates when Smith-Hansen, 2008). As these authors noted, it
treated with CBT vs. IPT. Unexpectedly, this inter- may be that the highly directive CBT relies, at least
active finding did not hold for the outcome of clini- partly, on patient communion/accommodation for
cally meaningful improvement in eating disorder quality alliance development.
symptoms more broadly. With regard to interperso- On other hand, the IPT therapist may hone in on
nal agency, as expected, we found that patients with the patient’s need for communion and close connec-
more problems of being overly agentic did better in tion (to the point of self-sacrifice and perhaps exploit-
IPT vs. CBT, whereas patients with more problems ability) as problematic. Although the IPT therapist
of being under agentic did better in CBT vs. IPT may conceptualize this interpersonal pattern as
for the outcome of clinically meaningful improve- central to the cause and maintenance of the eating
ment in broadly defined eating disorder symptoms. disorder (e.g., constantly trying to please critical
Unexpectedly, this interactive finding did not hold others by attempting to maintain a perceived ideal
for the purge frequency outcomes. Overall, the body shape), it may also threaten the patient who
results provided partial support for our hypotheses, fails to see attempts at close connection to others as
with no findings directly contradicting our expec- problematic. Such threat may prompt psychological
tations. Rather, the specific interpersonal trait by reactance and attempts to regain one’s sense of
treatment interactions had different effect patterns autonomy (in this case, in how they interact with
depending on the outcome variables examined, others), or a more pointed resistance to the direction
and we offer several plausible explanations why. of the IPT therapist, both of which can interfere with
Most simply, the fact that the hypothesized interac- treatment success (e.g., Aviram & Westra, 2011;
tive effects were only partially supported may have Beutler et al., 2011).
resulted from low statistical power. Supporting this For under communal/cold patients, the CBT
view, all findings that failed to reach statistical signifi- therapist’s bid for close collaboration around actively
cance were in the expected direction, and indeed changing thoughts and behaviors may prove too
often approached significance. Because the RCT threatening to the person motivated to keep distance
was not powered to test interactions (as is often the from others. And, again, with such threat could come
case), it is challenging to detect such effects that are heightened reactance or resistance, which could
often relatively small in size. It is possible that with undermine adaptive therapeutic process and gain in
a larger sample, we may have found the predicted this model. Moreover, cold/hostile interpersonal
interactive effects for both interpersonal problem traits have been negatively correlated with patient’s
dimensions on both outcomes. That being said, desire to focus their therapeutic work on the specific
because our findings were not uniformly significant, problems that bother them (CBT’s focus), and they
10 J. M. Gomez Penedo et al.

have been positively related to interpersonal concerns With regard to agency, as hypothesized, we found
during therapy (Gurtman & Balakrishnan, 1998); that for overly domineering patients, IPT vs. CBT
again, these are things that could derail a treatment’s is more likely to promote clinically meaningful
efficacy. Such heightened reactance or resistance for improvement in global eating disorder symptoms,
under communal patients may have resulted in whereas for overly submissive patients, CBT vs.
CBT being relatively less effective (compared to its IPT is more likely to promote such improvement.
general efficacy for patients with an average or high As expected, it may be that for BN patients with an
level of communality), rendering it equivalent to interpersonal style of, and need for, being overly
IPT for under communal patients. domineering toward others, IPT may be the more
Alternatively, or complementarily, the relative suitable treatment given that it may be less likely to
equivalence of the two treatments for reducing promote reactance or resistance than the more direc-
purge frequency rates in under communal patients tive CBT; that is, perhaps IPT’s focus on the inter-
could be driven by IPT therapists being better personal difficulties that underlie eating problems is
equipped to assess such coldness and hostility as a less threatening to overly domineering patients than
pattern central to eating pathology. And, without CBT’s explicit focus on changing disordered eating
needing to closely collaborate on strategies to cognitions and behaviors (Fairburn et al., 2003).
address explicitly eating pathology, the IPT thera- Moreover, IPT may be better situated to address
pist can instead address hostility and coldness directly problematic dominance as an explicit treat-
within an interpersonal formulation that a patient ment target, which could allow IPT therapists to
may be able to tolerate without imminent threat help patients more effectively navigate self and other
likely to promote negative process. Without such needs for agency in relationships (Wilson, 1996).
derailing negative processes, IPT could help With such interpersonal gains may come less risk
patients to develop new ways to deal with negative for disordered eating behavior. In contrast, for
affect, and to reduce the binge behaviors associated overly domineering patients, direct attempts by the
with it (Iacovino, Gredysa, Altman, & Wilfley, CBT therapist to reduce disordered eating cognitions
2012). Again, though, it is important to highlight and behavior may prompt patients to attempt to
that IPT was relatively more effective for under regain their autonomy. Such attempts could take
communal patients (relative to its general efficacy the form of psychological reactance or even more
for patients with an average or high level of com- pointed resistance to the direction of the CBT thera-
munality), but there was no evidence that it was pist, which can interfere with treatment success (e.g.,
more effective than CBT for under communal Aviram & Westra, 2011; Beutler et al., 2011).
patients. Additionally, whether the comparable effi- On the contrary, patients with an interpersonal
cacy of these two treatments for under communal style of (and need for) being overly submissive to
patients is due to CBT’s promotion of high reac- others may find the directive CBT therapist’s
tance rendering it less effective than usual, IPT’s symptom-focused approach suitable to their typical
ability to address these interpersonal problems style of relating to others. Consistent with this
directly without promoting resistance rendering it notion, these patients might be particularly likely to
more effective than usual, or a combination of these trustingly engage with CBT techniques and strat-
patterns remain unknown. egies, which has been associated with more positive
Moreover, it bears repeating that the communion by treatment outcomes in this approach (Kazantzis
treatment interaction did not replicate for the clinically et al., 2016). However, in IPT, such treatment com-
meaningful change outcome. It is possible that this pliance may lead the IPT therapist to overlook exces-
pattern of results emerged because purge frequency sively submissive behaviors as problematic. Given the
represents the type of outcome that is most targeted interpersonal focus of IPT, this could result in a focus
by CBT; that is, behavioral change on the overt and on other interpersonal content, which may limit the
objective number of times someone purges. Consist- potential efficacy of treatment due to a lack of focus
ent with this view, even when overly communal on the patients’ primary interpersonal difficulty of
patients are well-matched (and receive CBT), it may over submissiveness. This potential oversight could
not confer an advantage for the broader outcome result in poorer outcomes for overly submissive
that includes more subjective behavior. In other patients in IPT.
words, CBT’s advantage for this subgroup may be However, these results should be interpreted with
limited to the treatment’s most explicit target. some caution, as it is presently unknown why the
However, this notion requires testing, especially in agency by treatment interaction did not replicate for
larger samples; as noted, is possible that the predicted the purge frequency outcome. As noted, it is possible
interactive effect for clinically meaningful change that this could be due to power constraints. But, if
would emerge with greater statistical power. not, it will be important to understand what makes
 Psychotherapy Research 11

the problem of being overly communal a good match moderation analyses might see if differences in
with CBT for the very circumscribed issue of these putative mechanisms indeed account for why
purging, whereas being overly nonassertive in CBT traditional IPT and CBT perform better than the
results in a greater effect across the board on ED other under certain conditions.
symptoms. Additionally, future research would Several limitations characterize this study. First, as
need to investigate whether the beneficial influence noted, it is important to interpret the results with
of matching overly domineering patients to IPT also caution given that the interactive effects of interperso-
extends to more specific behavioral outcomes such nal problems by treatment were not uniform across
as purge frequency. Although speculative, it is poss- the two outcomes. Second, only self-report data
ible that the beneficial impact of IPT is most likely were available for assessing interpersonal functioning
to manifest as broad changes in eating disorder path- outside of the therapy room. Third, we only analyzed
ology, given its focus on addressing the underlying the differential symptomatic effects produced at post-
interpersonal issues of BN rather than on changing treatment. Fourth, the sample was restricted to
specific cognitive and behavioral targets. women. Although BN is much more prevalent
Such future work could also start focusing on more within women, 17% of diagnosed persons are men,
contemporary treatments, or iterations of CBT and which is a meaningful subsample (Hudson, Hiripi,
IPT, as the approaches in the present study are now Pope, & Kessler, 2007). Fifth, the therapists in the
somewhat dated. In the period since the Agras et al. study crossed conditions, and it is unknown if any
(2000) trial, there were several new treatments devel- preexisting allegiance or other biases may have
oped for BN. For example, Wonderlich et al. (2014) affected treatment processes or outcomes. Finally,
have created an integrative cognitive–affective the relatively low number of therapists in this study
therapy (ICAT) for BN. In addition to more tra- may have limited our ability to detect therapist
ditional cognitive strategies, ICAT also addresses effects on the outcomes.
psychological variables that might maintain bulimic Despite these limitations, to our knowledge, the
behaviors, such as change ambivalence, emotional present findings are the first to point to potential pre-
dysregulation, coping problems, interpersonal diffi- scriptive variables indicating for whom (among BN
culties, self-oriented behaviors (e.g., excessive self- patients) CBT or IPT (in their traditional forms) is
control), and self-oriented cognitive patterns (e.g., most likely to be effective for different outcomes.
self-discrepancy). The results also contribute to the broader literature
Regarding IPT for BN, we are aware of one new on interpersonal factors that moderate the compara-
development since the version used in the Agras tive treatment efficacy (Gomez Penedo, Constantino,
et al. (2000) trial. Murphy, Straebler, Basden, Coyne, Westra, & & Antony, 2017). Such work
Cooper, and Fairburn (2012) modified IPT to be a seems important given the growing perspective that
transdiagnostic treatment for eating disorders in it is the match between-patient characteristics and
general (IPT-ED). Therapists using this approach treatment type that may be primarily responsible for
are generally focused on patient’s current interperso- promoting and maintaining improvement (Beutler
nal relationships, rather than premorbid functioning, et al., 2011).
and they attempt to anticipate eventual problems
related to inevitable role transitions. Whether CBT-
based, IPT-based, or different altogether, future Notes
1
research should attempt to replicate the present find- Note that to establish the internal consistency of the global EDE
ings in these more contemporary approaches, both index, we used the same strategy as Constantino, Arnow, Blasey
for BN specifically and eating disorders in general. and Agras (2005) of calculating the alpha coefficient based on the
four subscales that combined to create the index.
Also with regard to future research, in addition to 2
Although Jacobson and Truax’s (1991) criteria are the most fre-
replication, it seems important to understand the quently used method for this purpose, it does not account for
specific mechanisms that may be responsible for regression toward the mean when measuring change (Lambert
one treatment’s superiority over another under a & Ogles, 2009; Speer & Greenbaum, 1995). Accounting for
such regression is particularly important in clinical trials, like
particular condition (e.g., for problems of interper-
the present one, that requires patients’ baseline symptoms to
sonal communion or agency). Although we con- exceed an established threshold of severity for inclusion in the
sidered that intrapsychic reactance, interpersonal trial (Johnson & Varghese, 1991).
3
resistance, and/or alliance development might For comparison, and following previous recommendations in the
explain the differential effects presented in CBT literature, we also calculated clinically meaningful change using
vs. IPT based on communal and agentic interperso- fully the Jacobson and Truax (1991) method. Based on this
approach, 32.7% of the sample were recovered (CBT = 37.3%,
nal problems, in this particular study, we did not IPT = 28.2%), 6.8% improved (CBT = 6.4%, IPT = 7.27%),
examine the potential mediating effect of these 60.0% unchanged (CBT = 56.4%, IPT = 63.6%), and 0.5%
common treatment factors. Further mediated deteriorated (CBT = 0.0%, IPT = 1.0%).
12 J. M. Gomez Penedo et al.
4
The interpersonal problems indices were grand mean centered randomized clinical trial for adults with symptoms of bulimia
prior to the creation of the interaction terms. nervosa. Journal of Consulting and Clinical Psychology, 84, 178–
5
As each interactive effect was tested in a separate model (one for 184.
each interpersonal predictor), we replicated these findings in a Agras, W. S., Walsh, T., Fairburn, C. G., Wilson, G. T., &
model including the main and interactive effects for both Kraemer, H. C. (2000). A multicenter comparison of cogni-
agency and communion as predictors of purge frequency. In tive-behavioral therapy and interpersonal psychotherapy for
this model, the interactive effect of communion by treatment bulimia nervosa. Archives of General Psychiatry, 57, 459–466.
remained significant (when predicting both the intercept and American Psychiatric Association. (1987). Diagnostic and statistical
linear slope). manual of mental disorders (3rd revised ed.). Washington, DC:
6
The result of this centering is that the main effect of treatment Author.
represents the relation between treatment group and weekly Ansell, E. B., Grilo, C. M., & White, M. A. (2012). Examining the
purge frequency change for patients with high and low values interpersonal model of binge eating and loss of control over
of the relevant baseline IIP-C index. Note that in these models eating in women. International Journal of Eating Disorders, 45,
only the main effects are impacted by the re-centering; the coef- 43–50. doi:10.1002/eat.20897
ficient for the interaction remains identical to the original model. Apple, R. F. (1999). Interpersonal therapy for bulimina nervosa.
7
Given the relatively high proportion of missing data, we also Journal of Clinical Psychology / in Session, 55, 715–725.
replicated our clinically meaningful change analyses using mul- Aviram, A., & Westra, H. A. (2011). The impact of motivational
tiple imputation via the Markov Chain Monte Carlo (MCMC) inter- viewing on resistance in cognitive–behavioural therapy
method in SPSS 23. First, we created 10 imputed datasets for for generalized anxiety disorder. Psychotherapy Research, 21,
the global EDE index at posttreatment. Second, we calculated 698–708. doi:10.1080/10503307.2011.610832
the average of the global EDE index for each participant across Beutler, L. E., Harwood, T. M., Michelson, A., Song, X., &
all imputed datasets. Third, we used that average to determine Holman, J. (2011). Resistance/reactance level. Journal of
whether patients met criteria for clinically meaningful change Clinical Psychology, 67, 133–142. doi:10.1002/jclp.20753
(i.e., patients were classified as recovered or not recovered). Constantino, M. J., Arnow, B. A., Blasey, C., & Agras, W. S.
Finally, we conducted two separate logistic regressions using (2005). The association between patient characteristics and
this variable as the criterion. Results of these analyses replicate the therapeutic alliance in cognitive – behavioral and interperso-
those of the main analyses, with a significant interactive effect nal therapy for bulimia nervosa. Journal of Consulting and Clinical
of treatment by agency on clinically meaningful improvement, Psychology, 73, 203–211. doi:10.1037/0022-006X.73.2.203
B = −0.138, SE = 0.07, Wald(1) = 4.423, p = .04, OR = .871, Constantino, M. J., Boswell, J. F., Bernecker, S. L., & Castonguay,
and no significant effect of treatment by communion on clinically L. G. (2013). Context-responsive psychotherapy integration as
meaningful improvement, B = 0.113, SE = 0.08, Wald(1) = a framework for a unified clinical science: Conceptual and
2.040, p = .16. empirical considerations. Journal of Unified Psychotherapy and
Clinical Science, 2, 1–20.
Constantino, M. J., & Smith-Hansen, L. (2008). Patient interper-
sonal factors and the therapeutic alliance in two treatments for
Supplementary data bulimia nervosa. Psychotherapy Research, 18, 683–698. doi:10.
1080/10503300802183702
Supplemental data for this article can be accessed 10.
Cooper, Z., & Fairburn, C. G. (1987). The eating disorder exam-
1080/10503307.2018.1425931. ination: A semi-structured interview for the assessment of the
specific psychopathology of eating disorders. International
Journal of Eating Disorders, 8, 1–8.
Disclosure statement Fairburn, C. G. (1997). Interpersonal psychotherapy for bulimia
nervosa. In D. M. Garner & P. E. Garfinkel (Eds.), Handbook
The authors report no conflicts of interest. of treatment for eating disorders (pp. 278–294). New York, NY:
Guilford.
Fairburn, C. G., Cooper, Z., & O’Connor, M. (2008). Eating dis-
order examination (16th ed.). In C. G. Fairburn (Ed.), Cognitive
Acknowledgement behavior therapy and eating disorders (pp. 265–308). New York,
NY: Guilford.
The authors are grateful to W. S. Agras, T. Walsh, Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive be-
C. G. Fairburn, and G. T. Wilson for allowing us haviour therapy for eating disorders: A “transdiagnostic” theory
access to the data set. and treatment. Behavior Research and Therapy, 41, 509–528.
doi:10.1016/S0005-7967(02)00088-8
Fairburn, C. G., Marcus, M. D., & Wilson, T. G. (1993).
Cognitive-behavioral therapy for binge eating and bulimia
ORCID nervosa: A comprehensive treatment manual. In C. G.
Fairburn & G. T. Wilson (Eds.), Binge eating: Nature, assess-
Samantha L. Bernecker http://orcid.org/0000-0002- ment, and treatment (pp. 361–404). New York, NY: Guilford.
8803-3311 Gomez Penedo, J. M., Constantino, M. J., Coyne, A. E., Westra,
H. A., & & Antony, M. M. (2017). Markers for context-respon-
siveness: Patient baseline interpersonal problems moderate the
efficacy of two psychotherapies for generalized anxiety disorder.
References
Journal of Consulting and Clinical Psychology, 85, 1000–1011.
Accurso, E. C., Wonderlich, S. A., Crosby, R. D., Smith, T. L., doi:10.1037/ccp0000233
Klein, M. H., Mitchell, J. E., … Peterson, C. B. (2016). Gurtman, M. B., & Balakrishnan, D. (1998). Circular measure-
Predictors and moderators of treatment outcome in a ment redux: The analysis and interpretation of interpersonal
 Psychotherapy Research 13
circle profiles. Clinical Psychology: Science and Practice, 5, 344– Muran, J. C., Segal, Z. V., Samstag, L. W., & Crawford, C. E.
360. (1994). Patient pretreatment interpersonal problems and thera-
Hartmann, A., Zeeck, A., & Barrett, M. S. (2010). Interpersonal peutic alliance in short-term cognitive therapy. Journal of
problems in eating disorders. International Journal of Eating Consulting and Clinical Psychology, 62, 185–190.
Disorders, 43, 619–627. doi:10.1002/eat.20747 Murphy, R., Straebler, S., Basden, S., Cooper, Z., & Fairburn, C.
Hay, P. J., & Claudino, A. M. (2010). Evidence-based treatment G. (2012). Interpersonal psychotherapy for eating disorders.
for the eating disorders. In W. S. Agras (Ed.), The Oxford hand- Clinical Psychology and Psychotherapy, 19, 150–158. doi:10.
book of eating disorders (p. 2010). Oxford: Oxford University 1002/cpp.1780
Press. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear
Hopwood, C. J., Clarke, A. N., & Perez, M. (2007). Pathoplasticity models: Applications and data analysis methods (2nd ed.).
of bulimic features and interpersonal problems. International Thousand Oaks, CA: Sage.
Journal of Eating Disorders, 40, 652–658. doi:10.1002/eat Rizvi, S. L., Peterson, C. B., Crow, S. J., & Agras, W. S. (2000).
Horowitz, L. M., Alden, L. E., Wiggins, J. S., & Pincus, A. L. Test-retest reliability of the eating disorder examination.
(2000). IIP: Inventory of interpersonal problems manual. San International Journal of Eating Disorders, 28, 311–316.
Antonio, TX: Psychological Corporation. Ruiz, M. A., Pincus, A. L., Borkovec, T. D., Echemendia, R. J.,
Horowitz, L. M., Rosenberg, S. E., Baer, B. A., Ureño, G., & Castonguay, L. G., & Ragusea, S. A. (2004). Validity of the
Villaseñor, V. S. (1988). Inventory of interpersonal problems: inventory of interpersonal problems for predicting treatment
Psychometric properties and clinical applications. Journal of outcome: An investigation with the Pennsylvania practice
Consulting and Clinical Psychology, 56, 885–892. research network. Journal of Personality Assessment, 83, 213–
Hudson, J. I., Hiripi, E., Pope, H. G., & Kessler, R. C. (2007). The 222. doi:10.1207/s15327752jpa8303_05
prevalence and correlates of eating disorders in the national Spangler, D. (1999). Cognitive-behavioral therapy for bulimia
comorbidity survey replication. Biological Psychiatry, 61, 348–358. nervosa: An illustration. Journal of Clinical Psychology / in
Iacovino, J. M., Gredysa, D. M., Altman, M., & Wilfley, D. E. Session, 55, 699–713.
(2012). Psychological treatments for binge eating disorder. Speer, D. C. (1992). Clinically significant change: Jacobson and
Current Psychiatry Reports 14, 432–446. doi:10.1007/s11920- Truax (1991) revisited. Journal of Consulting and Clinical
012-0277-8 Psychology, 60, 402–408.
Jacobson, N. S., & Truax, P. (1991). Clinical significance: A stat- Speer, D. C., & Greenbaum, P. E. (1995). Five methods for com-
istical approach to defining meaningful change in psychotherapy puting significant individual client change and improvement
research. Journal of Consulting and Clinical Psychology, 59, 12–19. rates: Support for an individual growth curve approach.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/ Journal of Consulting and Clinical Psychology, 63, 1044–1048.
2002127. Spitzer, R. L., Williams, J. B., Gibbon, M., & First, M. B. (1989).
Johnson, W. D., & Varghese, T. G. (1991). Effect of regression to Structured clinical interview for the DSM-III-R. Washington, DC:
the mean in the presence of within-subject variability. Statistics American Psychiatric Association.
in Medicine, 10, 1295–1302. Tanofsky-Kraff, M., & Wilfley, D. E. (2010). Interpersonal psy-
Kazantzis, N., Whittington, C., Zelencich, L., Kyrios, M., Norton, chotherapy for the eating disorders. In W. S. Agras (Ed.), The
P. J., & Hofmann, S. G. (2016). Quantity and quality of home- Oxford handbook of eating disorders (pp. 348–372). Oxford:
work compliance: A meta-analysis of relations with outcome in Oxford University Press.
cognitive behavior therapy. Behavior Therapy, 47, 755–772. Thompson-Brenner, H., Shingleton, R. M., Thompson, D. R.,
doi:10.1016/j.beth.2016.05.002 Satir, D. A., Richards, L. K., Pratt, E. M., & Barlow, D. H.
Klerman, G. L., Weismann, M. M., Rounsaville, B. J., & Chevron, (2016). Focused vs. broad enhanced cognitive behavioral
E. S. (1984). Interpersonal psychotherapy for depression. therapy for bulimia nervosa with comorbid borderline personal-
New York, NY: Basic Books. ity: A randomized controlled trial. International Journal of Eating
Lambert, M. J., & Ogles, B. M. (2009). Using clinical significance Disorders, 49, 36–49. doi:10.1002/eat.22468
in psychotherapy outcome research: The need for a common Wilson, G. T. (1993). Assessment of binge eating. In C. G.
procedure and validity data. Psychotherapy Research, 19, 493– Fairburn & G. T. Wilson (Eds.), Binge eating: Nature, assess-
501. doi:10.1080/10503300902849483 ment, and treatment (pp. 227–249). New York: Guilford Press.
Loeb, K. L., Pike, K. M., Walsh, B. T., & Wilson, G. T. (1994). Wilson, G. T. (1996). Treatment of bulimia nervosa: When CBT
The assessment of diagnostic features of bulimia nervosa: fails. Behavior Research and Therapy, 34, 197–212.
Interview versus self-report format. International Journal of Wilson, G. T., & Shafran, R. (2005). Eating disorders guidelines
Eating Disorders, 16, 75–81. from NICE. Lancet, 365, 79–81.
Loeb, K. L., Wilson, G. T., Labouvie, E., Pratt, E. M., Walsh, B. Wonderlich, S. A., Peterson, C. B., Crosby, R. D., Smith, T. L.,
T., Agras, W. S., & Fairburn, C. G. (2005). Therapeutic alli- Klein, M. H., Mitchell, J. E., & Crow, S. J. (2014). A random-
ance and treatment adherence in two interventions for bulimia ized controlled comparison of integrative cognitive-affective
nervosa : A study of process and outcome. Journal of therapy and cognitive-behavioral therapy-enhanced for
Consulting and Clinical Psychology, 73, 1097–1107. doi:10. bulimia nervosa. Psychological Medicine, 44, 543–553. doi:10.
1037/0022-006X.73.6.1097 1017/S0033291713001098.A