Available online at www.sciencedirect.

com

ScienceDirect
Behavior Therapy 56 (2019) 1063 –1074

www.elsevier.com/locate/bt

Baseline Client Interpersonal Agency Moderates the Indirect Effect

of Treatment on Long-term Worry in Variants of CBT for
Generalized Anxiety Disorder
Juan Martín Gómez Penedo
CONICET, Universidad de Buenos Aires
Michael J. Constantino
Alice E. Coyne
Felicia M. Romano
University of Massachusetts Amherst
Henny A. Westra
York University
Martin M. Antony
Ryerson University

clients with more vs. less problematic low agency at baseline,

In a recent trial for generalized anxiety disorder (GAD), CBT would still promote more in-session FS than MI-CBT,
cognitive-behavioral therapy (CBT) integrated with motiva- but this increase would in turn predict increased worry over
tional interviewing (MI) promoted more long-term worry follow-up. Clients receiving CBT (n = 43) or MI-CBT (n = 42)
reduction than CBT alone (Westra, Constantino, & Antony, rated their interpersonal problems at baseline and their worry
2016). A follow-up analysis found that CBT vs. MI-CBT after treatment and across 12-month follow-up. Therapists
clients evidenced greater increases in friendly submissiveness rated clients’ in-session FS multiple times. As predicted,
(FS) across treatment, which in turn promoted lower long- multilevel modeling revealed that for clients with more
term worry (Constantino, Romano, Coyne, Westra, & problematic low agency, CBT vs. MI-CBT facilitated greater
Antony, 2018). It was unsurprising that traditional directive FS, which in turn related to increased worry across follow-up.
CBT promoted more FS than when person-centered MI was For clients with more problematic high agency, CBT’s
integrated; however, given that problematic low agency facilitation of greater FS related to reduced worry across
characterizes GAD, that greater FS promoted better outcome follow-up. A baseline interpersonal problem characteristic of
was unexpected. To further unpack this unexpected result, we GAD may have implications for treatment matching and for
tested the following moderated mediation hypothesis: for appreciating different pathways to long-term improvement,
or deterioration, for different GAD subgroups.

This research was supported by the Canadian Institutes of Health Keywords: interpersonal agency; in-session interpersonal behavior;
Research (MOP -114909). generalized anxiety disorder; cognitive-behavioral therapy; motiva-
Address correspondence to Juan Martín Gómez Penedo, Ph.D., tional interviewing
Secretaria de Investigaciones, Facultad de Psicología, Universidad
de Buenos Aires, 2353 Lavalle St., Ciudad de Buenos Aires,
Argentina; e-mail: [email protected]. A GROWING BODY OF RESEARCH has linked interper-
0005-7894/© 2019 Association for Behavioral and Cognitive Therapies. sonal difficulties with generalized anxiety disorder
Published by Elsevier Ltd. All rights reserved. (GAD). For example, compared with nonanxious
1064 g ó m e z p e n e d o e t a l .

controls, persons with GAD are more likely to Consistent with this notion, a recent clinical trial
perceive interpersonal threat in their interactions for GAD compared traditional CBT to CBT that
(Mogg, Mathews, & Eysenck, 1992; Mogg, Millar, integrated motivational interviewing (MI) specifi-
& Bradley, 2000), have greater relational hyper- cally to address clients’ interpersonal resistance to
vigilance and suspiciousness, be more easily the treatment or therapist (Westra, Constantino,
offended by others, be more dependent on others, & Antony, 2016). In the context of GAD, resistance
and seek more reassurance from others (Gasperini, may result from clients’ ambivalence about relin-
Battaglia, Diaferia, & Bellodi, 1990). Furthermore, quishing their worry (despite that being the primary
interpersonal concerns are a primary worry domain target of CBT), which is often perceived as both
in GAD (Breitholtz, Johansson, & Öst, 1999; a source of distress and as a way to maintain
Craske, Rapee, Jackel, & Barlow, 1989; Hoyer, control (Newman, Llera, Erickson, Przeworski, &
Becker, & Roth, 2001; Sanderson & Barlow, Castonguay, 2013). Moreover, given that clients
1990), and people with vs. without this disorder with GAD tend to present as under agentic and
are more likely to have insecure attachments overly communal, resistance may represent an
(Cassidy, Lichtenstein-Phelps, Sibrava, Thomas, adaptive interpersonal risk-taking attempt at asser-
& Borkovec, 2009; Mickelson, Kessler, & Shaver, tiveness. In these moments, it is possible that typical
1997; Viana & Rabian, 2008). CBT strategies, such as continuing to push for worry
Although interpersonal problems in GAD can reduction and reinforcing the treatment rationale,
vary (Przeworski et al., 2011), several studies have might unintentionally recapitulate the interpersonal
demonstrated a prototypical pattern of self-reported problems that are most characteristic of GAD (i.e., a
excessive friendliness and submissiveness, such as person with GAD submits to an interacting other’s
being nonassertive, exploitable, and overly nurtur- demands; Westra & Constantino, in press). Ulti-
ing with others (Salzer et al., 2008; Salzer, Pincus, mately, if such an interpersonal pattern is indeed tied
Winkelbach, Leichsenring, & Leibing, 2011). Fur- to the etiology and maintenance of GAD pathology,
ther, such problematic “friendly submissiveness” is such a pattern might render traditional CBT less
more common in clients with GAD than for those effective for these clients.
with other psychiatric disorders (Gomez Penedo, In contrast, MI is a client-centered approach that
Constantino, Coyne, Westra, & Antony, 2017). seeks to enhance clients’ self-efficacy and assertiveness
Given the potential centrality of these interpersonal (Miller & Rollnick, 2002), precisely during moments
problems in the etiology and maintenance of GAD, it when they resist (Aviram, Westra, Constantino, &
follows that they may also influence the process and Antony, 2016). In these moments, the MI-CBT
outcome of treatments for this condition (Borkovec, therapists in the Westra et al. (2016) trial engaged a
Ray, & Stober, 1998; Newman, Castonguay, “spirit” of empathy, evocation, and collaboration,
Borkovec, Fisher, & Nordberg, 2008). while simultaneously using techniques designed to
Empirically supporting this view, several studies elicit clients’ own arguments for change. Such a
in which clients received the most commonly response might represent a novel, corrective experi-
delivered interventions for GAD, behavioral and ence for clients with GAD who may be used to others
cognitive-behavioral therapy (CBT), suggest that dominating and dismissing their needs (Constantino
more interpersonal difficulties are associated with & Westra, 2012). Over time, this emancipating
poorer outcomes both at posttreatment and long-term behavior on the part of the therapist might promote
follow-up, and that many clients complete treatment an increase in clients’ self-confidence and assertiveness
with unresolved interpersonal problems (Borkovec, during treatment, which could translate to intrinsi-
Newman, Pincus, & Lytle, 2002; Millstein, Orsillo, cally motivated actions to reduce worry. Moreover,
Hayes-Skelton, & Roemer, 2015). Moreover, where- any increased assertiveness might eventually translate
as CBT has well-established efficacy for anxiety to other relationships outside of therapy, which could
disorders in general (Watts, Turnell, Kladnitski, mitigate interpersonal risk for GAD symptoms.
Newby, & Andrews, 2015), it has more modest Supporting these notions, Westra and colleagues
success rates for GAD specifically (Cuijpers et al., (2016) found that although CBT and MI-CBT
2014; Hunot, Churchill, Teixeira, & Silva De Lima, achieved comparable posttreatment outcomes, MI-
2007), which implicates interpersonal problems as CBT significantly outperformed CBT across a 1-year
a specific risk factor for poorer response. Thus, some follow-up. Specifically, whereas CBT clients’ worry
have argued that explicitly integrating relational either remained stable or slightly increased across the
strategies into CBT for GAD could represent one follow-up period, MI-CBT clients showed continued
pathway to enhancing its efficacy (Borkovec et al., worry reduction. The authors posited that this
2002; Newman et al., 2008; Westra & Constantino, “sleeper” effect may have emerged, in part, because
in press). the benefits of the corrective relational experience of
 baseline client interpersonal agency 1065

receiving support for autonomy-taking (and the but we expected that the association between FS and
increased self-efficacy it may promote) may not long-term worry reduction would differ depending
emerge until after the therapist is no longer available, on the degree to which clients’ are characteristically
when clients are left to implement worry-reduction under agentic when they begin treatment (agency as a
strategies on their own. moderator of the “b” path).
To test this notion, Constantino et al. (2018)
examined whether an increase in clients’ interperson-
Method
participants
al assertiveness and a decrease in their submissiveness Clients
during therapy explained the additive long-term Eighty-five adults were randomly assigned to either
efficacy of MI-CBT vs. CBT. The authors used the MI-CBT (n = 42) or CBT (n = 43) across two sites in
Impact Message Inventory (IMI; Kiesler & Schmidt, Toronto, Canada. To be included in the parent trial,
2006) to examine changes in clients’ friendly clients were required to have a principal diagnosis of
dominant and friendly submissive interpersonal GAD based on Diagnostic and Statistical Manual
impacts on their therapist during treatment. Although of Mental Disorders version IV, Text Revision
an increase in friendly dominance (FD) was related to (DSM-IV-TR; American Psychiatric Association,
lower long-term worry levels across both treatments, 2000) and version 5 (DSM- 5; American
such changes did not mediate the comparative Psychiatric Association, 2013) criteria. Clients also
treatment effect across follow-up. Thus, the authors had to score equal to, or above, the cutoff of 68
suggested that an increase in clients’ FD might be a points for high worry severity, as measured by the
facilitative factor common to both CBT and MI-CBT. Penn State Worry Questionnaire (PSWQ; Meyer,
On the other hand, change in friendly submissiveness Miller, Metzger, & Borkovec, 1990). Clients taking
(FS) did mediate the relation between treatment antidepressant medications were eligible if they (a)
condition and worry level, but in the opposite used the same medication and dosage for at least 3
direction than expected. Namely, although CBT months prior to the start of the trial, and (b) agreed to
clients demonstrated the hypothesized greater in- maintain the same medication and dose throughout.
creases in FS throughout treatment compared to Clients who discontinued a psychopharmacological
MI-CBT clients, such increases were associated with treatment were included if they had a 3-month
lower worry at 12-month follow-up. Put another washout period before the trial. To enhance gener-
way, CBT’s facilitation of greater increases in FS alizability, the only comorbidity exclusions were
allowed these clients to achieve lower long-term psychotic spectrum disorders, bipolar disorders,
worry levels that became closer to, though were still major cognitive impairment, substance dependence
slightly higher than, the MI-CBT clients’ long-term within the past 6 months, and/or significant current
worry levels. The authors speculated that increased suicidal ideation.
FS related to better outcome in CBT because it may Table 1 presents the sample characteristics at
reflect a type of trusting compliance with this directive baseline by treatment group. Regarding demo-
therapeutic approach. graphic characteristics, clients in the two conditions
However, as per the interpersonal context dis- differed only on gender, χ 2 (1) = 4.24, p = .04,
cussed previously, greater FS, or compliance, should with CBT having more women and fewer men
theoretically reinforce, not help to reduce, GAD than MI-CBT. Additionally, there were more clients
symptoms. Thus, it is possible that the extent to on psychotropic medications in CBT vs. MI-CBT,
which increased FS is beneficial may depend on the χ 2 (1) = 3.94, p = .05. Furthermore, as measured by
degree to which clients are characteristically under the Change Questionnaire (CQ; Miller & Johnson,
agentic in their relationships. In the present study, 2008), CBT clients presented with greater motiva-
we tested this moderated mediation question by tion to change than MI-CBT clients, t(83) = 2.55,
again drawing on the Westra et al. (2016) trial data. p = .01. Finally, medication use and motivation
Specifically, we hypothesized that CBT would also varied between the two treatment sites. Specific
promote greater increases in client FS than MI-CBT to the present study, the treatments did not sig-
(a direct effect already established in the aforemen- nificantly differ on baseline worry (PSWQ) or on
tioned Constantino et al., 2018, study), which problematic interpersonal agency, as per the
would in turn predict better follow-up outcome for Inventory of Interpersonal Problems-Circumplex
clients who are less problematically under agentic (IIP-C; Horowitz, Alden, Wiggins, & Pincus, 2000)
(submissive) at baseline and worse follow-up out- discussed below.
come for clients who are more problematically under
agentic (submissive) at baseline. In other words, prior Therapists
research has established that CBT vs. MI-CBT would Twenty-one female therapists self-selected into
promote greater increases in client FS (the “a” path), treatment condition to control for allegiance and
1066 g ó m e z p e n e d o e t a l .

Table 1
Baseline Client’s Characteristics by Treatment Condition
Variables CBT (n = 43) MI-CBT (n = 42)
M (SD) n % M (SD) n %
Age 34.19 (11.92) 32.45 (10.54)
Gender ⁎
Women 41 95.34 34 80.95
Men 2 4.65 8 19.05
Race
White 32 74.42 31 73.81
Asian 5 11.62 6 14.29
Hispanic 2 4.65 1 2.38
Multiracial/other 4 9.30 4 9.52
Annual household income
Less than 25,000 10 23.26 6 14.29
25,000-50,000 9 20.93 8 19.05
50,000-75,000 11 25.58 8 19.05
75,000-100,000 8 18.60 6 14.29
100,000 5 11.63 13 30.95
Education
High school or less 4 9.30 2 4.76
Some college/university 13 30.23 9 21.43
Completed college 18 41.86 19 45.24
Some graduate school 8 18.60 12 28.57
Marital status
Single 19 44.19 18 42.86
Cohabiting/married 23 54.76 24 57.14
Psychotropic use ⁎
Yes 14 32.56 6 14.29
No 29 67.44 36 85.71
Comorbidity
Anxiety disorder 31 72.09 29 69.05
Depression/dysthymia 17 39.53 13 30.95
PSWQ 75.05 (3.43) 74.69 (3.44)
CQ ⁎ 107.23 (8.76) 101.60 (11.50)
IIP Agency -3.46 (2.63) -3.73 (1.98)
Note. M = Mean; SD = standard deviation; CBT = cognitive-behavioral therapy; MI = motivational interviewing; PSWQ = Penn State Worry
Questionnaire; CQ = Change
Questionnaire; IIP = Inventory of Interpersonal Problems
⁎ There is a significant difference between the two conditions (p b .05)

crossover effects; 9 provided MI-CBT and 12 pro- Treatments

vided CBT. On average, across both conditions, the All clients received 15 weekly (50 min) sessions,
therapists were 28.76 years old (SD = 3.46 years) plus two “booster” sessions at 1 and 3 months after
and had 294.74 hours of clinical experience (SD = termination. Clients in MI-CBT received up to 4
420.44 hours). Neither age nor experience differed initial sessions of MI only followed by 11 sessions
between treatment conditions. For both condi- of CBT with MI responsively integrated to address
tions, pre-trial training included participating in client resistance. (Although most MI-CBT clients
clinical workshops and treating pilot cases with received these 4 initial sessions of pure MI, for those
feedback provided by experts of the respective clients who were highly motivated to engage in
therapies. During the trial, therapists received change-oriented strategies, the shift into integrative
weekly supervision by these same experts. As MI-CBT occurred 1-2 sessions earlier.)
expected with this additive design, independent
observers rated therapists in both conditions as CBT
showing a high level of CBT competence. Also, as Therapists delivered CBT according to several
expected, observers rated MI use and integrity evidence-based protocols for GAD (e.g., Craske &
higher in MI-CBT vs. CBT alone (see Westra et al., Barlow, 2006; Zinbarg, Craske, & Barlow, 2006).
2016 for details). The main techniques included psychoeducation,
 baseline client interpersonal agency 1067

cognitive self-monitoring and restructuring, progres- nonassertive + .71 (intrusive + vindictive – socially
sive muscle relaxation, behavioral interventions, inhibited – exploitable)]. The possible scores for
and, when necessary, sleep strategies. CBT therapists this dimension range from -9.68 (under agentic)
were trained to use strategies both for preventing and to 9.68 (overly agentic). The six subscales used to
responding to client noncompliance (e.g., resistance) compute this dimension demonstrated adequate
that were drawn from the CBT literature (e.g., Beck, internal consistency in the present sample (Cronbach’s
2005). Specifically, preventative strategies included αs = .72 to .85).
collaborative negotiation of goals and the provision
Interpersonal Impacts
of a rationale for all treatment tasks, whereas
To assess clients’ interpersonal style during therapy,
responsive strategies included problem solving or
therapists completed the Impact Message Inventory
engaging in a functional analysis.
(IMI; Kiesler & Schmidt, 2006). The IMI relies
MI-CBT on the theory that the interpersonal style of a given
Therapists delivered MI-CBT according to the same person can be assessed by the impact messages
CBT protocols, with MI techniques integrated received by an interacting other during an exchange
according to Westra’s (2012) guidelines for GAD. (Kiesler, 1996). This idea is based on Kiesler’s
MI is a client-centered approach that seeks to complementary principle, which postulates that
resolve clients’ change ambivalence and to reduce the positions of two interactants tend to be similar
any resulting behavioral resistance to the treatment in terms of affiliation (e.g., friendliness pulls for
(Miller & Rollnick, 2002). MI consists of specific a friendliness), but tend to be opposed in terms
techniques such as “rolling with” vs. challenging of the control or power distribution in the dyad
client resistance and developing discrepancies (e.g., dominance pulls for submission).
between clients’ current and most valued selves. Also a circumplex, the IMI consists of 56 items
Therapists also work from a foundational “spirit” rated on a scale from 1 (not at all) to 4 (very much
of evocation, empathy, and support for client so), which divides into eight subscales that reflect
autonomy (rather than acting as an external change combinations of the interpersonal dimensions of
agent or trying to explicitly counteract resistance). affiliation (ranging from hostile to friendly impacts)
In the Westra et al. (2016) trial, the integration and control (ranging from submissive to dominant
of MI into CBT was accomplished in two ways. impacts): hostile, hostile-dominant, dominant,
First, CBT strategies were delivered with MI “spirit.” friendly-dominant, friendly, friendly-submissive,
Second, therapists fully shifted into MI (and submissive, hostile-submissive. In this study, we
suspended CBT) when they observed markers of focused on the friendly-submissive scale based on
client resistance; when such resistance was resolved, the aforementioned evidence that changes in FS
therapists shifted back into CBT. mediated the long-term effect of treatment on worry
(Constantino et al., 2018). To calculate FS at
each time point, we used the following weighted
measures
formula: FS + .707 (F + S). This weighted score has
Interpersonal Problems
a theoretical range from 16.90 to 67.59, with
To assess interpersonal problems, clients completed
higher scores representing more FS. Previous
the 32-item IIP-C (Horowitz et al., 2000), a
validation studies found that the IMI possesses
measure that has been found to possess good
adequate internal consistency, good convergent and
psychometric properties (Soldz, Budman, Demby,
discriminant validity, and adequate temporal sta-
& Merry, 1995). The items, which are rated on a
bility (Schmidt, Wagner, & Kiesler, 1999). In the
scale from 0 (not at all) to 4 (extremely), are
current study, Cronbach’s alphas for the FS vector
theoretically distributed in a circumplex structure,
ranged from .80 to .94.
based on the primary interpersonal dimensions
of communion (ranging from overly cold to Worry
overly nurturant behaviors) and agency (ranging To assess worry, clients completed the PSWQ (Meyer
from overly nonassertive to overly domineering et al., 1990), which includes 16 items rated from 1
behaviors). Across these two dimensions, the IIP-C (not at all typical of me) to 5 (very typical of me). The
yields eight interpersonal problem domains (i.e., PSWQ total score has a theoretical range of 16 to 80,
[being too] domineering, intrusive, overly nurtur- with higher scores representing greater worry. The
ant, exploitable, nonassertive, socially inhibited, PSWQ possesses good internal consistency and test-
cold, and vindictive). For the present study, we retest reliability, as well as good concurrent, discrim-
focused on the agency dimension, which is calcu- inant, and convergent validity (Brown, Antony, &
lated according to the following weighted formula Barlow, 1992; Meyer et al., 1990). In this study, the
(Ruiz et al., 2004): Agency = .25 [domineering – PSWQ total score demonstrated excellent internal
1068 g ó m e z p e n e d o e t a l .

consistency across posttreatment and follow-up Finally, for our primary analyses, we conducted a
measurements (Cronbach’s αs = .96 to .97). multilevel moderated mediation model (Preacher,
Rucker, & Hayes, 2007) using the Mplus 8.1
PROCEDURE program (Muthén & Muthén, 1998-2017). For this
See Westra et al. (2016) for a full description of the mediational analysis, we used the Bayesian estima-
flow of participants through the trial. For variables tor (Muthén & Asparouhov, 2012), as indirect
pertinent to the present study, clients completed effects are typically not normally distributed.
the IIP-C at baseline, and the PSWQ at baseline, Bayesian analysis does not assume normality and
posttreatment, and at 6- and 12-month follow up. instead provides the entire distribution for estimates
The therapists completed the IMI after Sessions 1, (known as posterior distributions), which allows
5, 9, 13, and 15. All study procedures (as well as for the generation of credible intervals based on
the secondary analysis of the deidentified dataset) these posterior distributions. A 95% credible
received institutional review board approval at the interval (CI) indicates that there is a 95% chance
two data collection sites. that the interval contains the true estimate of the
effect. Thus, a credible interval for an estimate that
DATA ANALYSES does not contain zero can be used as an approxi-
First, as per Constantino et al. (2018), we created mation of a more traditional frequentist signifi-
our change in FS mediator variable by conducting cance test.
a 2-level linear growth model with within-client As previous research on the Westra et al. (2016)
change at level 1 and between client differences at trial has established that there were no therapist
level 2 (Raudenbush & Bryk, 2002). Although effects on worry, we fit a two-level model with
there was another level of nesting in the dataset repeated measures of worry (level 1) nested within
(clients within therapists), we did not include clients (level 2). Specifically, this model tested an
therapists as a third level of analysis given upper level mediation with the effect of a level-2
Constantino et al.’s (2018) finding that therapists predictor (treatment condition; MI-CBT = 0; CBT =
explained b 1% of the variability in change in FS 1) on a level 1 outcome variable (worry) mediated
across treatment. From the 2-level model, we by a level-2 variable (change in FS; i.e. 2 → 2 → 1
output empirical Bayes (EB) estimates of the weekly mediation; Bauer, Preacher, & Gil, 2006). Note
rate of change in FS during treatment (see the that we focused solely on worry change during
Supplemental Material for a full description of the the follow-up period due to our interest in better
equation for this model). These EB estimates are understanding the differential treatment effect that
weighted by their reliability; that is, when clients emerged at this time in the main trial (but not over
had missing IMI data, their change scores were acute treatment; Westra et al., 2016). We also
weighted toward the group mean. This approach focused on follow-up to fulfill the methodological
allows the model to generate estimates for all clients requirement of maintaining temporal precedence
with at least one completed IMI measurement (thus, between the predictor, mediator, and outcome
the full sample was retained for this variable). variable when testing mediation (Kraemer,
However, it is important to note that only 18% of Kiernan, Essex, & Kupfer, 2008).
clients had any missing IMI data. Regarding the Then, we incorporated another level-2 variable in
other two measures used in this study, only 16% the model (client’s baseline interpersonal problems
had any missing PSWQ data, and there were no of agency) as a moderator of the b path of the
missing baseline IIP data. mediation effect. In these analyses at level-1 (within-
Second, to create our outcome variable, we client effects), we estimated worry change during
followed the same approach as Westra et al. follow-up by regressing the PSWQ scores on time
(2016) and residualized out the effects of the (in weeks), centered at the end of follow-up. Thus,
aforementioned variables that differed between this model estimated each client’s weekly follow-up
the conditions at baseline; that is, as baseline worry change (slope) and their worry level at
motivation to change and medication status signif- 12 months (intercept). At level-2 (between-client
icantly varied both between treatment conditions effects), these variables dropped down to become
and the site where the therapy was conducted, we outcome variables; that is, each client’s follow-up
first residualized out the effect of site on motivation worry change and worry level at 12-months were
to change and medication status. Then, we resi- regressed on treatment condition (c’ path), FS
dualized out the effect of motivation and medica- change (b path), baseline FS, baseline interpersonal
tion (with site now removed) on PSWQ scores. The problems of agency, and the interaction of change
subsequent analyses were conducted based on these in FS and interpersonal problems of agency (b
residualized PSWQ scores. path moderator). Also at level-2, change in FS was
 baseline client interpersonal agency 1069

regressed on treatment condition (a path). Finally, agency, the interactive effect of change in FS and
to test the overall significance of the moderated agency, and treatment group. Similarly, the main
mediational pathway, we calculated Indexes of effect of baseline agency on worry change (γ03 =
Moderated Mediation (Hayes, 2015). These index- 0.70, SE = 0.39, CI95 [-0.02, 1.48]) and 12-month
es were computed by multiplying the coefficient worry (γ03 = -0.41, SE = 0.33, CI95 [-1.05, 0.21])
of the “a” path by the coefficients of the regression were not significant, controlling for change in FS,
of client’s follow-up worry change and worry level the interactive effect of change in FS and agency,
at 12-months on the interaction of the mediator and treatment group.
(change in FS) and the “b” path moderator However, as expected, we found that baseline
(interpersonal problems of agency). See the Online agency significantly moderated the effect of change in
Supplement for the multilevel equations for this FS on long-term worry change (γ14 = -3.37, SE =
model. 1.60, CI95 [-6.55, -0.26]). See Figure 1 for a graphical
depiction of all paths in this moderated mediation
Results model. Further, the index of moderated mediational
P R E L I M I NA R Y A N A L Y S E S testing the extent to which the effect of treatment
Results of the multilevel FS model indicated that, on follow-up worry change through change in FS
on average, FS increased by 0.19 units per week varied as a function of baseline problems of agency
(p b .001). Moreover, results indicated that clients was significant (β = -0.24, SE = 0.18, CI95 [-0.68,
significantly varied in their weekly change in FS, -0.01]). As depicted in Figure 2, for more problem-
u1j = 0.07; χ 2(76) = 189.41, p b .001, suggesting atically under agentic clients (1 SD below the grand
that between-client differences in this change could mean of agency), increased FS was associated with
explain variability in client outcomes. As noted, increased worry during follow-up, whereas for
we output EB estimates representing each client’s problematically over agentic clients (1 SDs above the
weekly FS change during treatment for use as our grand mean of agency), increased FS was associated
mediator variable in our primary analyses. with reduced worry.
To further understand this finding, we calculated
P R I M A R Y A NA L Y S E S simple slopes and the region of significance for
Results of our multilevel moderated mediational a three-way interaction (time x FS x agency) in a
analysis showed that treatment condition was hierarchical linear model, using Preacher, Curran,
significantly related to FS change during therapy and Bauer’s (2006) online calculator. The results
(γ11 = .08, SE = .03, CI95 [0.02, 0.15]), with CBT of the regions of significance analysis revealed
clients experiencing a greater average increase in FS that clients who were 1 SD above the mean in
during treatment than MI-CBT clients. The main agency (problematically over agentic) experienced
effects of change in FS during treatment on worry significant worry reduction when they evidenced a
change across follow-up (γ02 = 5.04, SE = 4.19, during-treatment weekly rate of FS increase that
CI95 [-3.27, 13.65]) and worry level at 12-months was between 0.01 and 0.47 units. Clients who were
(γ12 = -19.61, SE = 13.51, CI95 [-47.72, 5.87]) were 1 SD below the agency mean (problematically
not significant, controlling for baseline problematic under agentic) experienced no change in worry, or

FIGURE 1 Moderated mediational effects of treatment (MI-CBT = 0, CBT = 1; X) on worry

change across follow-up (Y) through weekly changes in FS (M), with the effect of change in FS
moderated by clients’ baseline agentic interpersonal problems (W).
Note. FS = Friendly-submissiveness; Δ = Change
* Bayesian 95% credible interval does not include zero.
1070 g ó m e z p e n e d o e t a l .

FIGURE 2 Effects of the weekly changes in FS during treatment on the weekly rate of change
in worry across follow-up, moderated by client’s baseline agency. Over agentic and under agentic
interpersonal problems were defined as ± 1 SD (2.32) from the sample’s agency M (-3.59)
at baseline.
Note. PSWQ = residualized scores from the Penn State Worry Questionnaire; FS = IMI friendly
submissiveness scores; FU = follow-up.

an increase in worry, when their weekly rate of clients, greater increases in FS (again, as facilitated
FS increase was within the range of 0.27 to 3.57 more by CBT than MI-CBT) related to decreased
units. These under agentic clients only experienced worry across follow up, whereas for under agentic
significant worry reduction when they evidenced clients, greater increases in FS (again, as facilitated
weekly reductions in FS or just a small weekly more by CBT than MI-CBT) related to increased
increase in FS (up to 0.27 units). These findings worry across 12-month follow up. However, this
support our hypothesis that greater increases in FS moderated indirect effect was not replicated for
will be beneficial for over agentic clients, whereas endpoint worry level at the 12-month assessment.
reductions or just small increases in FS will be Although speculative, it is possible that the rates of
beneficial for clients who are under agentic. change in worry associated with the moderated
In contrast, the change in FS by agentic interper- mediation effect are relatively slow (compared to,
sonal problems interactive effect did not signifi- for example, what one might see during treatment
cantly predict 12-month worry level (γ04 = -6.42, when a therapist is actively trying to reduce clients’
SE = 5.03, CI95 [-16.49, 3.51]). Additionally, the worry), and that a longer-term follow-up period
index of moderated mediation was not significant would have been needed to detect a significant
(β = -0.47, SE = 0.52, CI95 [-1.82, 0.30]). Finally, moderated mediation effect on final worry level.
when controlling for change in FS, initial FS level, This potential design limitation notwithstanding
baseline agency, and the interactive effect of change these results shed further light on the clients with
in FS and agency, there was still a significant effect GAD for whom CBT, through its candidate change
of treatment; that is, MI-CBT vs. CBT clients mechanism of increased FS, might be more or less
experienced greater worry reduction across follow- beneficial. It is plausible, and perhaps even likely, that
up (γ11 = 3.81, SE = 1.44, CI95 [0.84, 6.60]) and for overly agentic persons who may be accustomed
had lower 12-month worry (γ01 = 11.24, SE = 4.33, to dominating others and who usually tend to be
CI95 [2.71, 19.06]). Supplemental Table 1 presents distrusting in their relationships (Gurtman, 1992;
all parameters estimated in this model. Horowitz, 2004), learning to trust and rely on
another’s direction over the course of treatment
Discussion might represent a novel and corrective relational
We tested whether GAD clients’ baseline interper- experience (Constantino & Westra, 2012). Over
sonal agency problems moderated the indirect effect time, clients’ increased trust and reliance on the CBT
of treatment on long-term worry through during- therapist’s direction might not only result in lowered
treatment changes in clients’ FS impacts on their worry, but could also translate to a reduced need to
therapist. As prior research had already established get their interpersonal needs met through domineer-
(Constantino et al., 2018), CBT vs. MI-CBT clients ing behavior. If this experience in therapy is salient
experienced greater increases in FS. However, the enough, it might even translate to other relationships
present study demonstrated that for over agentic outside of therapy, which could improve the balance
 baseline client interpersonal agency 1071

in, and quality of, such relationships (Coyne et al., with GAD based on their particular interpersonal
2018). Such improved relational quality could even features (Beutler, Someah, Kimpara, & Miller,
become a protective factor against the recurrence of 2016). Such decision-making would fit with the call
GAD symptoms. for a more personalized medicine in evidence-based
On the other hand, for under agentic clients, mental health care in general (DeRubeis et al., 2014),
increased FS (as facilitated by CBT) may be more and for GAD clients specifically (Etchebarne, Juan,
harmful than helpful. For these individuals, becom- & Roussos, 2016). Moreover, the identification of
ing more deferent to the directive CBT therapist baseline interpersonal problems as an empirically
might unintentionally infix, or even worsen, their supported marker for treatment assignment fits with
characteristic excessive submissiveness (Horowitz, an evidence-based responsiveness model known
2004). Given the link between this type of interper- as context-responsive psychotherapy integration
sonal problem and the etiology and maintenance (Constantino, Boswell, Bernecker, & Castonguay,
of GAD symptomatology (Borkovec et al., 2002; 2013). This model proposes an if/then approach
Newman et al., 2008), it might also contribute to whereby treatment adaptations are prompted by
increased worry over the long term. the presence of empirically supported markers or
Thus, consistent with previous findings (Gomez contexts. In the present case, if GAD clients have
Penedo et al., 2017), overly submissive clients may baseline problems of being overly agentic, then
benefit from the autonomy-granting nature of MI. standard CBT may lead to interpersonal experiences
Specifically, when the MI-CBT therapist empathic- that help reduce these problems, which in turn
ally supports and validates clients’ self-efficacy and facilitates long-term worry reduction. In contrast, if
experience, especially when they take an atypical GAD clients present with baseline problems of being
risk to resist the treatment’s direction, under agentic under agentic, then integrating MI into CBT may
clients may begin to take more agency in this safe help foster beneficial reductions in submissiveness,
and trustworthy dyad. Over time, this exchange which in turn facilitates long-term worry reduction.
could become a different type of novel, corrective Moreover, once treatment begins (after person-
experience (Constantino & Westra, 2012; Westra alized assignment), therapists can monitor clients’
& Constantino, in press); in this case, learning that interpersonal behaviors, such as with the IMI, and
the MI-CBT therapist (a new important other) draw on our present region of significance analyses
values and encourages the client’s assertiveness to inform further responsiveness. For example,
could reduce his or her pathological deference. suppose a client presents as problematically over
Eventually, such changes could translate to other agentic, and a therapist proceeds to administer
relationships outside of therapy (Newman et al., standard CBT. If upon tracking FS change during
2008; Wachtel, 2014), thereby lowering this GAD treatment the client experiences a moderate increase
risk factor and facilitating long-term worry reduc- in FS, then the clinician may be wise to stay the CBT
tion and durability of treatment gain. course (as our data would suggest that this client is
Additionally, it is worth noting that although the on track for long-term worry reduction). However,
present results suggest potential benefits of therapist if the client experiences either any reduction in FS
responsivity to both over and under agentic clients, (a movement consistent with their problematically
for the present sample, it appears most important over agentic style) or a very large increase in FS
that therapists do not unintentionally recapitulate a (perhaps too extreme of a movement to the other
pattern of dominance with an overly submissive end of the problematic agency spectrum), then the
client. In other words, the beneficial impact of MI- therapist may be wise to alter their treatment course
CBT vs. CBT for clients with problematic low agency to address the client’s interpersonal patterns that
was more pronounced than the beneficial impact are posing risk for long-term nonresponse or harm.
of CBT vs. MI-CBT for overly agentic clients. Given On the other hand, suppose a client presents as
that the present sample tended to be very under problematically under agentic, and a therapist
agentic on average, it is possible that the beneficial proceeds to administer integrative MI-CBT. If upon
impact of CBT vs. MI-CBT for overly agentic clients tracking FS change during treatment the client
would have been more pronounced had our sample experiences any decrease in FS, or FS remains
included more clients with truly problematic high relatively stable, then the clinician may be wise to
agency (rather than clients who simply had less stay this integrative course (as our data would suggest
problematic low agency). Future research will need that this client is on track for long-term worry
to test this possibility in other samples with greater reduction). However, if the client experiences an
variability in presenting client agency. increase in FS beyond the “cut-point” we found in the
Clinically, the present results may have implica- region of significance analyses, then the therapist may
tions for systematic treatment selection for clients be wise to alter their approach, such as by employing
1072 g ó m e z p e n e d o e t a l .

other strategies that might encourage the client’s interpersonal processes during therapy can be more
assertiveness and reduce his submissiveness. or less beneficial (or perhaps even harmful) based on
The present study had several limitations. First, a client’s specific presenting interpersonal problem.
clients’ during-therapy interpersonal behavior was Thus, it represents one humble step toward answer-
not measured directly, but rather inferred through ing Gordan Paul’s iconic question for psychother-
complementarity theory. Thus, there might be bias apy process research: “What treatment, by whom, is
in therapist ratings associated, for example, with most effective for this individual with that specific
social desirability or acquiescence. Further studies problem, and under which set of circumstances?”
might benefit from using an observer-coding system (Paul, 1967, p. 11).
to evaluate interpersonal processes and comple- Conflict of Interest Statement
mentarity in client and therapist dyads. Such a The authors declare that there are no conflicts of interest.
method could yield a more detailed picture of the
psychotherapy interaction, clients’ interpersonal Appendix A. Supplementary data
changes over time, and how these processes relate Supplementary data to this article can be found
to treatment characteristics, process variables, and online at https://doi.org/10.1016/j.beth.2019.01.007.
therapeutic change. Second, we did not measure
therapist interpersonal style. Relatedly, although we References
assumed that CBT was more directive than MI-CBT,
we did not include an empirical measure of such American Psychiatric Association. (2000). Diagnostic and
directiveness in our analyses. However, it is worth statistical manual of mental disorders (4th ed., text revision).
noting that, based on adherence coding (using Arlington, VA: American Psychiatric Press.
observer-based measures that assess inherent differ- American Psychiatric Association. (2013). Diagnostic and
ences in agenda-driven/directive actions [CBT] vs. statistical manual of mental disorders (5th ed.). Arlington,
VA: American Psychiatric Press.
client-centered stances [MI]), CBT therapists were Aviram, A., Westra, H. A., Constantino, M. J., & Antony,
significantly more directive than MI-CBT therapists M. M. (2016). Responsive management of early resistance
throughout treatment. Third, the results were based in cognitive-behavioral therapy for generalized anxiety
on random assignment of participants to treatment disorder. Journal of Consulting and Clinical Psychology,
condition. To fully substantiate the hypothesis that 84, 783–794. https://doi.org/10.1037/ccp0000100
Bauer, D. J., Preacher, K. J., & Gil, K. M. (2006). Conceptu-
assigning clients to treatment (CBT or MI-CBT) alizing and testing random indirect effects and moderated
based on their presenting interpersonal difficulties mediation in multilevel models: New procedures and
would improve clinical outcomes, experimental trials recommendations. Psychological Methods, 11, 142–163.
for which such an assignment process is implemented https://doi.org/10.1037/1082-989X.11.2.142
are needed. Fourth, the relatively small number of Beck, A. T. (2005). The current state of cognitive therapy: A
40-year retrospective. Archives of General Psychiatry, 62,
therapists (N = 21) could have limited our ability 953–959. https://doi.org/10.1001/archpsyc.62.9.953
to detect meaningful therapist effects on both the Beutler, L. E., Someah, K., Kimpara, S., & Miller, K. (2016).
mediator (change in FS) and the outcome (worry Selecting the most appropriate treatment for each client.
change). Fifth, this study’s primary research ques- International Journal of Clinical and Health Psychology, 16,
tion was based on the results of other studies that 99–108. https://doi.org/10.1016/j.ijchp.2015.08.001
Borkovec, T. D., Newman, M. G., Pincus, A. L., & Lytle, R.
used the same dataset, which implies a risk of (2002). A component analysis of cognitive-behavioral therapy
sample-specific results that could have a limited for generalized anxiety disorder and the role of interpersonal
generalization to the broader population of clients problems. Journal of Consulting and Clinical Psychology, 70,
with GAD. Future research should replicate these 288–298. https://doi.org/10.1037//0022-006X.70.2.288
findings in new samples of GAD clients treated with Borkovec, T. D., Ray, W. J., & Stober, J. (1998). Worry: A
cognitive phenomenon intimately linked to affective, phys-
CBT vs. MI-CBT. Finally, as mentioned above, the iological, and interpersonal behavioral processes. Cognitive
moderated mediation effect was not significant Therapy and Research, 22, 561–576. https://doi.org/10.
for worry level at 12-months. Although speculative, 1023/A:1018790003416
it is possible that the follow-up rate of change in Breitholtz, E., Johansson, B., & Öst, L. G. (1999). Cognitions
worry is relatively slow (compared to the rates of in generalized anxiety disorder and panic disorder clients.
A prospective approach. Behaviour Research and Therapy,
symptom change observed during acute treatment), 37, 533–544. https://doi.org/10.1016/S0005-7967(98)
so it may take more time for between-group 00147-8
differences in worry level to emerge (i.e., beyond Brown, T. A., Antony, M. M., & Barlow, D. H. (1992).
12-months posttreatment). Therefore, future re- Psychometric properties of the Penn State Worry Question-
search should evaluate worry change over a more naire in a clinical anxiety disorders sample. Behaviour
Research and Therapy, 30, 33–37. https://doi.org/10.1016/
prolonged follow-up period. 0005-7967(92)90093-V
Beyond these limitations, this study provides Cassidy, J., Lichtenstein-Phelps, J., Sibrava, N. J., Thomas,
further evidence about the ways in which specific C. L., & Borkovec, T. D. (2009). Generalized anxiety disorder:
 baseline client interpersonal agency 1073

connections with self-reported attachment. Behavior Therapy, Hayes, A. F. (2015). An index and test of linear moderated
40, 23–38. https://doi.org/10.1016/j.beth.2007.12.004 mediation. Multivariate Behavioral Research, 50, 1–22.
Constantino, M. J., Boswell, J. F., Bernecker, S. L., & https://doi.org/10.1080/00273171.2014.962683
Castonguay, L. G. (2013). Context-responsive psychother- Horowitz, L. M. (2004). Interpersonal foundations of psycho-
apy integration as a framework for a unified clinical science: pathology. Washington, DC: American Psychological
Conceptual and empirical considerations. Journal of Unified Association.
Psychotherapy and Clinical Science, 2, 1–20. Horowitz, L. M., Alden, L. E., Wiggins, J. S., & Pincus, A. L.
Constantino, M. J., Romano, F. M., Coyne, A. E., Westra, H. A., (2000). IIP: Inventory of interpersonal problems manual.
& Antony, M. M. (2018). Client interpersonal impacts San Antonio, TX: Pearson Assessment.
as mediators of long-term outcome in cognitive-behavioral Hoyer, J., Becker, E. S., & Roth, W. T. (2001). Characteristics of
therapy integrated with motivational interviewing for worry in gad clients, social phobics, and controls. Depression
generalized anxiety disorder. Psychotherapy Research, 28, and Anxiety, 13, 89–96. https://doi.org/10.1002/da.1021
861–872. https://doi.org/10.1080/10503307.2017. Hunot, V., Churchill, R., Teixeira, V., & Silva De Lima, M.
1301689 (2007). Psychological therapies for generalised anxiety
Constantino, M. J., & Westra, H. A. (2012). An expectancy- disorder (Review). Cochrane Database of Systematic Reviews,
based approach to facilitating corrective experiences in 2007, 1, CD001848. https://doi.org/10.1002/14651858.
psychotherapy. In C. E. Hill, & L. G. Castonguay (Eds.), cd001848.pub3
Transformation in psychotherapy: Corrective experiences Kiesler, D. J. (1996). Contemporary interpersonal theory and
across cognitive behavioral, humanistic, and psychodynamic research: Personality, psychopathology, and psychotherapy.
approaches (pp. 121–139). Washington, DC: American Hoboken, NJ: John Wiley and Sons.
Psychological Association. Kiesler, D. J., & Schmidt, J. A. (2006). The impact message
Coyne, A. E., Constantino, M. J., Gómez Penedo, J. M., Gnall, inventory-circumplex (IMI-C) manual. Redwood City, CA:
K. E., McBride, C., & Ravitz, P. (2018). The relation Mind Garden.
between changes in participants’ interpersonal impact Kraemer, H. C., Kiernan, M., Essex, M., & Kupfer, D. J.
messages and outcome in interpersonal psychotherapy (2008). How and why criteria defining moderators and
for depression. Journal of Psychotherapy Integration, 28, mediators differ between the Baron & Kenny and
475–488. https://doi.org/10.1037/int0000125 MacArthur approaches. Health Psychology, 27, 101–108.
Craske, M. G., & Barlow, D. H. (2006). Mastery of your https://doi.org/10.1037/0278-6133.27.2(suppl.).s101
anxiety and worry: Workbook (2nd ed.). New York, NY: Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D.
Oxford University Press. (1990). Development and validation of the Penn State
Craske, M. G., Rapee, R. M., Jackel, L., & Barlow, D. H. Worry questionnaire. Behaviour Research and Therapy, 28,
(1989). Qualitative dimensions of worry in DSM-III-R 487–495. https://doi.org/10.1016/0005-7967(90)90135-6
generalized anxiety disorder subjects and nonanxious Mickelson, K. D., Kessler, R. C., & Shaver, P. R. (1997). Adult
controls. Behaviour Research and Therapy, 27, 397–402. attachment in a nationally representative sample. Journal
https://doi.org/10.1016/0005-7967(89)90010-7 of Personality and Social Psychology, 73, 1092–1106.
Cuijpers, P., Sijbrandij, M., Koole, S., Huibers, M., Berking, https://doi.org/10.1037//0022-3514.73.5.1092
M., & Andersson, G. (2014). Psychological treatment Miller, W. R., & Johnson, W. R. (2008). A natural language
of generalized anxiety disorder: A meta-analysis. Clinical screening measure for motivation to change. Addictive
Psychology Review, 34, 130–140. https://doi.org/10.1016/j. Behaviors, 33, 1177–1182. https://doi.org/10.1016/j.
cpr.2014.01.002 addbeh.2008.04.018
DeRubeis, R. J., Cohen, Z. D., Forand, N. R., Fournier, J. C., Miller, W. R., & Rollnick, S. (2002). Motivational interviewing:
Gelfand, L. A., & Lorenzo-Luaces, L. (2014). The person- Preparing people for change. (2nd ed.). New York, NY:
alized advantage index: Translating research on prediction Guilford Press.
into individualized treatment recommendations. A demon- Millstein, D. J., Orsillo, S. M., Hayes-Skelton, S. A., & Roemer,
stration. PLoS ONE, 9, e83875. https://doi.org/10.1371/ L. (2015). Interpersonal problems, mindfulness, and therapy
journal.pone.0083875 outcome in an acceptance-based behavior therapy for
Etchebarne, I., Juan, S., & Roussos, A. J. (2016). El abordaje generalized anxiety disorder. Cognitive Behaviour Therapy,
clínico de la preocupación en el trastorno de ansiedad 44, 491–501. https://doi.org/10.1080/16506073.2015.
generalizada por parte de terapeutas cognitivo conductuales 1060255
de Buenos Aires [Clinical approach to worry in generalized Mogg, K., Mathews, A., & Eysenck, M. (1992). Attentional bias
anxiety disorder by cognitive-behavioral therapists from to threat in clinical anxiety states. Cognition & Emotion, 6,
Buenos Aires]. Anuario Anxia, 22, 6–20. 149–159. https://doi.org/10.1080/02699939208411064
Gasperini, M., Battaglia, M., Diaferia, G., & Bellodi, L. (1990). Mogg, K., Millar, N., & Bradley, B. P. (2000). Biases in eye
Personality features related to generalized anxiety disorder. movements to threatening facial expressions in generalized
Comprehensive Psychiatry, 31, 363–368. https://doi.org/10. anxiety disorder and depressive disorder. Journal of
1016/0010-440x(90)90044-s Abnormal Psychology, 109, 198–204. https://doi.org/10.
Gomez Penedo, J. M., Constantino, M. J., Coyne, A., & 1037//0021-843X
Westra, H. y Antony M. (2017). Markers for context- Muthén, B., & Asparouhov, T. (2012). Bayesian SEM: A more
responsiveness: Client baseline interpersonal problems flexible representation of substantive theory. Psychological
moderate the efficacy of two psychotherapies for generalized Methods, 17, 313–335. https://doi.org/10.1037/a0026802
anxiety disorder. Journal of Consulting and Clinical Muthén, L. K., & Muthén, B. O. (1998-2017). Mplus (8 th ed.).
Psychology, 85, 1000–1011. https://doi.org/10.1037/ Los Angeles, CA: Muthen & Muthen.
ccp0000233 Newman, M. G., Castonguay, L. G., Borkovec, T. D., Fisher, A. J.,
Gurtman, M. B. (1992). Trust, distrust, and interpersonal & Nordberg, S. S. (2008). An open trial of integrative therapy
problems: A circumplex analysis. Journal of Personality and for generalized anxiety disorder. Psychotherapy: Theory,
Social Psychology, 62, 989–1002. https://doi.org/10.1037// Research, Practice, Training, 45, 135–147. https://doi.org/
0022-3514.62.6.989 10.1037/0033-3204.45.2.135
1074 g ó m e z p e n e d o e t a l .

Newman, M. G., Llera, S. J., Erickson, T. M., Przeworski, A., & 588–591. https://doi.org/10.1097/00005053-199009000-
Castonguay, L. G. (2013). Worry and generalized anxiety 00006
disorder: A review and theoretical synthesis of evidence on Schmidt, J. A., Wagner, C. C., & Kiesler, D. J. (1999).
nature, etiology, mechanisms, and treatment. Annual Review Psychometric and circumplex properties of the Octant
of Clinical Psychology, 9, 275–297. https://doi.org/10.1146/ Scale Impact Message Inventory (IMI-C): A structural
annurev-clinpsy-050212-185544 evaluation. Journal of Counseling Psychology, 46, 325–334.
Paul, G. L. (1967). Strategy of outcome research in psycho- https://doi.org/10.1037//0022-0167.46.3.325
therapy. Journal of Consulting Psychology, 31, 109–118. Soldz, S., Budman, S., Demby, A., & Merry, J. (1995). A
https://doi.org/10.1037/h0024436 short form of the Inventory of Interpersonal Problems
Preacher, K. J., Curran, P. J., & Bauer, D. J. (2006). Circumplex Scales. Assessment, 2, 53–63. https://doi.org/10.
Computational tools for probing interaction effects in multiple 1177/1073191195002001006
linear regression, multilevel modeling, and latent curve Viana, A. G., & Rabian, B. (2008). Perceived attachment:
analysis. Journal of Educational and Behavioral Statistics, Relations to anxiety sensitivity, worry, and GAD symptoms.
31, 437–448. https://doi.org/10.3102/10769986031004437 Behaviour Research and Therapy, 46, 737–747. https://doi.org/
Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing 10.1016/j.brat.2008.03.002
moderated mediation hypotheses: Theory, methods, and Wachtel, P. L. (2014). An integrative relational point of
prescriptions. Multivariate Behavioral Research, 42, 185–227. view. Psychotherapy, 51, 342–349. https://doi.org/10.1037/
https://doi.org/10.1080/00273170701341316 a0037219
Przeworski, A., Newman, M. G., Pincus, A. L., Kasoff, M. B., Watts, S. E., Turnell, A., Kladnitski, N., Newby, J. M., &
Yamasaki, A. S., Castonguay, L. G., & Berlin, K. S. (2011). Andrews, G. (2015). Treatment-as-usual (TAU) is anything
Interpersonal pathoplasticity in individuals with generalized but usual: A meta-analysis of CBT versus TAU for anxiety and
anxiety disorder. Journal of Abnormal Psychology, 120, depression. Journal of Affective Disorders, 175, 152–167.
286–298. https://doi.org/10.1037/a0023334 https://doi.org/10.1016/j.jad.2014.12.025
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear Westra, H. A. (2012). Motivational interviewing in the
models: Applications and data analysis methods (2nd ed.). treatment of anxiety. New York, NY: Guilford Press.
Thousand Oaks, CA: Sage. Westra, H. A., & Constantino, M. J. (in press). Integrative
Ruiz, M. A., Pincus, A. L., Borkovec, T. D., Echemendia, R. J., psychotherapy for generalized anxiety disorder. In J. C.
Castonguay, L. G., & Ragusea, S. A. (2004). Validity of the Norcross, & M. R. Goldfried (Eds.), Handbook of psychother-
inventory of interpersonal problems for predicting treatment apy integration (3rd ed.). New York, NY: Oxford University
outcome: an investigation with the Pennsylvania Practice Press.
Research Network. Journal of Personality Assessment, 83, Westra, H. A., Constantino, M. J., & Antony, M. M. (2016).
213–222. https://doi.org/10.1207/s15327752jpa8303_05 Integrating motivational interviewing with cognitive-behavioral
Salzer, S., Pincus, A. L., Hoyer, J., Kreische, R., Leichsenring, F., therapy for severe generalized anxiety disorder: An allegiance-
& Leibing, E. (2008). Interpersonal subtypes within gener- controlled randomized clinical trial. Journal of Consulting and
alized anxiety disorder. Journal of Personality Assessment, Clinical Psychology, 84, 768–782. https://doi.org/10.1037/
90, 292–299. https://doi.org/10.1080/00223890701885076 ccp0000098
Salzer, S., Pincus, A. L., Winkelbach, C., Leichsenring, F., & Zinbarg, R. E., Craske, M. G., & Barlow, D. H. (2006). Mastery
Leibing, E. (2011). Interpersonal subtypes and change of of your anxiety and worry: Therapist guide (2nd ed.).
interpersonal problems in the treatment of clients with New York, NY: Oxford University Press.
generalized anxiety disorder: A pilot study. Psychotherapy,
48, 304–310. https://doi.org/10.1037/a0022013
Sanderson, W. C., & Barlow, D. H. (1990). A description R E C E I V E D : September 30, 2018
of clients diagnosed with DSM-III-R generalized anxiety A C C E P T E D : January 30, 2019
disorder. Journal of Nervous and Mental Disease, 178, A V A I L A B L E O N L I N E : February 10, 2019