Behaviour Research and Therapy 50 (2012) 469e478
Contents lists available at SciVerse ScienceDirect
Behaviour Research and Therapy
journal homepage: www.elsevier.com/locate/brat
Longitudinal treatment mediation of traditional cognitive behavioral therapy and
acceptance and commitment therapy for anxiety disorders
Joanna J. Arch a, *, Kate B. Wolitzky-Taylor b, Georg H. Eifert c, Michelle G. Craske b
a
University of Colorado Boulder, Department of Psychology and Neuroscience, 345 UCB Muenzinger, Boulder, CO 80309-0345, USA
University of California, Los Angeles, Department of Psychology, 405 Hilgard Avenue, Los Angeles, CA 90095, USA
c
Chapman University, Department of Psychology, One University Drive, Orange, CA 92866, USA
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 12 August 2011
Received in revised form
20 April 2012
Accepted 23 April 2012
Objective: To assess the relationship between session-by-session putative mediators and treatment
outcomes in traditional cognitive behavioral therapy (CBT) and acceptance and commitment therapy
(ACT) for mixed anxiety disorders.
Method: Session-by-session changes in anxiety sensitivity and cognitive defusion were assessed in 67
adult outpatients randomized to CBT (n ¼ 35) or ACT (n ¼ 32) for a DSM-IV anxiety disorder.
Results: Multilevel mediation analyses revealed significant changes in the proposed mediators during
both treatments (p < .001, d ¼ .90e1.93), with ACT showing borderline greater improvements than CBT
in cognitive defusion (p ¼ .05, d ¼ .82). Anxiety sensitivity and cognitive defusion both significantly
mediated post-treatment worry; cognitive defusion more strongly predicted worry reductions in CBT
than in ACT. In addition, cognitive defusion significantly mediated quality of life, behavioral avoidance,
and (secondary) depression outcomes across both CBT and ACT (p < .05, R2 change ¼ .06e.13), whereas
anxiety sensitivity did not significantly mediate other outcomes.
Conclusions: Cognitive defusion represents an important source of therapeutic change across both CBT
and ACT. The data offered little evidence for substantially distinct treatment-related mediation pathways.
Ó 2012 Elsevier Ltd. All rights reserved.
Keywords:
Cognitive behavioral therapy (CBT)
Acceptance and commitment therapy (ACT)
Anxiety
Mediation
Cognitive defusion
Anxiety sensitivity
Determining if and how psychotherapies work is a central task of
clinical science. Thus far, the field has succeeded greatly at
demonstrating that specific therapies are effective for specific
disorders, but we are still working to demonstrate “how.”
Addressing how therapies work requires identifying the specific
variables that drive change within specific treatments, or treatment
mediators.
Anxiety disorders, the most common class of psychiatric disorders (Kessler, Berglund, Demler, Jin, & Walters, 2005), are characterized by fear, anxiety, and behavioral avoidance. Over the past few
decades, cognitive behavioral therapy (CBT) has become the most
empirically supported psychotherapy for anxiety disorders (see
Craske, 2010). Meta-analyses confirm the efficacy of CBT for the
treatment of anxiety disorders relative to wait-list, expectancy and
attention control conditions (Butler, Chapman, Forman, & Beck,
2006; Hofmann & Smits, 2008) and psychodynamic therapy
(Tolin, 2010). However, relatively few studies rigorously examine
* Corresponding author. Tel.: þ1 303 492 4634; fax: þ1 303 492 2967.
E-mail addresses: Joanna.Arch@Colorado.edu (J.J. Arch), ktaylor@psych.ucla.edu
(K.B. Wolitzky-Taylor), geifert@chapman.edu (G.H. Eifert), Craske@psych.ucla.edu
(M.G. Craske).
0005-7967/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.brat.2012.04.007
mediators of CBT for anxiety disorders. Understanding CBT-specific
vs. treatment-common processes requires comparing CBT mediators to those of another active psychotherapy (see Arch & Craske,
2008; Kraemer, Wilson, Fairburn, & Agras, 2002). Yet, relatively
few studies compare CBT for anxiety disorders to fully active
psychotherapy-based treatments (Tolin, 2010). Of those that do, very
few compare treatment mediation between the two approaches.
Acceptance and commitment therapy (ACT; Hayes, Strosahl, &
Wilson, 1999, 2012), an acceptance-based behavioral therapy, has
been applied specifically to the treatment of anxiety disorders (e.g.,
Arch et al., in press; Eifert & Forsyth, 2005; Eifert et al., 2009). With
roots in the behavioral and experiential therapy traditions, ACT
cultivates mindfulness, acceptance and cognitive defusion with the
aim of decreasing avoidance of internal discomfort, increasing
psychological flexibility, and above all, promoting behavior change
consistent with personal values (Hayes et al., 1999). Case studies,
multiple-baseline studies, and an initial randomized study (Twohig
et al., 2010) provide nascent evidence that ACT is an effective
treatment for anxiety disorders, including obsessive compulsive
disorder (Twohig, Hayes, & Masuda, 2006; Twohig et al., 2010),
social anxiety disorder (Dalrymple & Herbert, 2007), panic disorder
(Eifert et al., 2009), and posttraumatic stress disorder (Orsillo &
470
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
Batten, 2005). In addition, an acceptance-based behavioral therapy
for generalized anxiety disorder has shown effectiveness relative to
a wait-list control condition (Roemer, Orsillo, & Salters-Pedneault,
2008). We recently completed a randomized trial comparing ACT
and CBT for the treatment of a mixed anxiety disorder sample,
showing similar effectiveness of ACT compared to traditional CBT
(Arch et al., in press). Thus, initial evidence suggests that ACT is
effective for the treatment of anxiety disorders.
Within the context of anxiety disorder treatment, the question
remains of how traditional CBT (herein abbreviated as CBT) and ACT
work and whether they work for the same reasons. The cognitive
model for CBT identifies reductions in anxiety-related thoughts and
beliefs as the central treatment process, leading to subsequent fear
or symptom reduction (see Craske, 2010). Although the evidence
has been mixed, several CBT studies report that changes in cognitions predict or mediate symptom improvements. For example,
Hofmann (2004) found that pre- to post-treatment reductions in
negative social cost ratings (patient ratings of “how bad would it be”
if a feared social event occurred) predicted symptom outcomes in
CBT for social phobia, though such changes also mediated outcomes
in a comparison behavioral therapy condition. Smits, Rosenfeld,
McDonald, and Telch (2006) established that session-by-session
estimates for probability of a feared social outcome predicted
subsequent fear reduction in social phobia. Further, Hofmann et al.
(2007) demonstrated that changes in panic-related cognitions
across pre-post-follow up predicted panic disorder symptom
improvement in conditions that included CBT (CBT-only, CBT plus
medication) but not in a medication-only condition. Meuret,
Rosenfield, Hofmann, Seidel, and Bhaskara (2010) demonstrated
changes in negative beliefs about panic and anxiety-related sensations were bidirectionally related to panic symptom severity in
cognitive therapy but not in capnometry-assisted respiratory
training for panic disorder. Elsewhere, pre- to post-treatment
reductions in fear of and negative beliefs about panic and anxietyrelated sensations accounted for significant variance in panic
disorder symptom change in CBT (Smits, Powers, Cho, & Telch,
2004). These studies provide initial evidence for a significant relationship between cognitive and symptom changes in CBT for panic
disorder and social phobia, and suggest that this relationship is
sometimes specific to CBT. However, with two notable exceptions
(Meuret et al., 2010; Smits et al., 2006), these studies did not assess
changes in the mediator variable during treatment, and thus did not
establish the temporal precedence of the mediator required for full
mediation testing. Furthermore, few compared mediation of CBT to
that of another active psychotherapy.
Whereas the cognitive model of CBT posits that change in
thoughts and beliefs about anxiety predicts symptom reductions,
ACT posits that “buying into” or being fused with anxiety-related
thoughts and beliefs is a core cause of disordered anxiety (Eifert &
Forsyth, 2005; Forsyth, Eifert, & Barrios, 2006). Consequently, ACT
aims to reduce fusion with anxiety-related thoughts through
“cognitive defusion” or flexibly distancing from the literal meaning of
cognition so that cognition no longer rigidly dictates behavior.
Acceptance is employed for approaching rather than avoiding
anxiety-related thoughts, feelings, and physical sensations. A small,
multiple-baseline study of ACT for social phobia (Dalrymple &
Herbert, 2007) established that pre- to mid-treatment changes on
a measure of cognitive defusion, acceptance, and valued action predicted subsequent symptom improvement. A randomized trial
comparing ACT to relaxation training for obsessive compulsive
disorder (Twohig et al., 2010) found greater pre- to post-treatment
increases on a measure of cognitive defusion, acceptance, and
value-guided action in ACT than in relaxation training, although the
differences were no longer significant at 3 month follow up. Two
small studies in samples of psychotic inpatients (e.g., Gaudiano &
Herbert, 2006a,b; see Hayes, Luoma, Bond, Masuda, & Lillis, 2006)
and depressed outpatients (Zettle & Hayes, 1986), support the notion
that changes in the believability of distressing hallucinations or
cognitions (a measure of cognitive defusion) predict outcomes in ACT.
Collectively, these studies identify cognitive defusion, acceptance, and value-guided action as likely mediators of ACT for
anxiety disorders. Furthermore, evidence (Twohig et al., 2010)
suggests that changes in these variables may be specific to ACT at
post-treatment, at least compared to relaxation training. The
evidence to date is only preliminary, however, and the methodological quality varies widely. For example, only one study
(Dalrymple & Herbert, 2007) evaluated the temporal precedence of
the mediator and none compared ACT mediators to those of CBT.1
In summary, few previous studies fulfill the highest standards
for tests of treatment mediation. Full testing of treatment mediation requires that the mediation measure be administered during
treatment, and preferably at multiple time points during treatment,
to establish the temporal precedence of the mediator relative to
outcome measures (Kraemer et al., 2002; MacKinnon, Fairchild, &
Fritz, 2007). Also, the mediation measure should be treatmentspecific and psychometrically valid. Methodologies such as multilevel modeling provide greater statistical power and permit finegrained analyses such as the time sequencing of effects (Kenny,
Kashy, & Bolger, 1998). Several exemplary studies in the anxiety
disorders literature fulfill these standards (e.g., Meuret, Rosenfield,
Hofmann, Suvak, & Roth, 2009; Moscovitch, Hofmann, Suvak, & InAlbon, 2005) but address different mediation questions than does
the present study. Our aim is to apply these rigorous standards to
investigating treatment mediation of CBT and ACT for anxiety
disorders.
Our discussion has focused more on cognitive mediation
(broadly defined) of CBT and ACT for two reasons. First, extant
studies of mediation in CBT and ACT have focused on cognitive
mediation, namely, reductions of anxiety-related beliefs in CBT
(thought content) and reductions of cognitive fusion or believability
in ACT (thought context). Second, we believe that cognitive mediators offered greater potential to distinguish CBT and ACT than
behavioral mediators. In that CBT and ACT are both behaviorallybased therapies, they share the goal of reducing behavioral avoidance. ACT focuses on engaging in valued behaviors that have been
avoided due to life narrowing that results from attempts to control
symptoms, whereas CBT focuses on engaging in behaviors that have
been avoided due to their association with fear and anxiety. In
practice, however, these approaches may result in similar reductions in behavioral avoidance, which would likely drive change in
both treatments (see Arch & Craske, 2008). For these two reasons,
we focused the present investigation on assessing cognitive mediation of CBT and ACT, defined in the broad sense of mediators that
reflect either the context (how we relate to) or content of cognition.
Our study investigated two central questions: 1) Do CBT and ACT
affect the theorized mediators for each treatment, showing greater
reductions in beliefs about the harmful effects of anxiety (i.e.
anxiety sensitivity) in CBT and greater increases in cognitive
defusion in ACT? 2) Do changes in anxiety sensitivity and cognitive
defusion mediate treatment outcomes? Specifically, do treatmentspecific processes mediate outcomes within the specified treatment only (anxiety sensitivity mediates CBT but not ACT outcomes,
whereas cognitive defusion mediates ACT but not CBT outcomes),
or, alternatively, do treatment-specific processes mediate outcomes
across both treatments (anxiety sensitivity and cognitive defusion
mediate outcomes across both CBT and ACT)?
1
For effectiveness studies of ACT vs. CBT in undiagnosed patient samples, see
Forman, Herbert, Moitra, Yeomans, & Geller, 2007; Lappalainen et al., 2007.
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
Methods
Participants
Sixty-seven adult outpatients (Ps) meeting criteria for one or
more DSM-IV anxiety disorders were included in the mediation
analyses, including Ps randomized to CBT (n ¼ 35) and ACT (n ¼ 32).
Ps were included if they completed at least 6 sessions of psychotherapy (half of treatment) and at least one session-by-session
mediation measure during treatment (see Measures) in order to
provide a sufficient foundation for investigating therapy process.
Seven Ps (10%) who completed 6 or more sessions did not finish
treatment; they were included in analyses of mediator change
during treatment but not in post-treatment outcome analyses.
Thus, the n’s for post-treatment outcome analyses are lower than
for change in mediators during treatment.2
Ps were recruited from the greater West Los Angeles area; 51%
were female, 63% were Caucasian, 13% were Latino/HispanicAmerican, 12% were Asian-American, 6% were African-American, 1%
were Native American, and 4% were biracial. The mean age was 37
years (11.8 SD, range 19e60) with 15 years of education (2.1 SD,
range 9e21) and 22% were married, 55% were single, 10% were
cohabiting, and 12% were separated or divorced. CBT and ACT Ps
did not differ significantly on any demographic variables, all
ps > .3. Clinical diagnoses were ascertained by highly trained
doctoral students and professional research assistants using the
Anxiety Disorders Interview Schedule-IV-Revised (ADIS-IV-R)
(DiNardo, Brown, & Barlow, 1994). Principal diagnoses included 36%
(24/67) with panic disorder with or without agoraphobia, 24% (16/
67) with social anxiety disorder, 19% (13/67) with generalized
anxiety disorder, 13% (9/67) with obsessive compulsive disorder, 4%
with specific phobia (3/67) and 3% with posttraumatic stress
disorder (2/67). Of these, 11% (7/66)3 additionally met criteria for
major depression. CBT and ACT Ps did not differ significantly on
principal diagnoses. Although CBT (45.7%, 16/35) had more Ps with
principal PD/A than ACT (25.0%, 8/32), this difference did not reach
significance, c2(1) ¼ 3.12, p ¼ .08. Diagnostic assessments and
treatment took place at the Anxiety Disorders Research Center at
the University of California Los Angeles (UCLA), Department of
Psychology.
Exclusion criteria were active suicidal ideation, very severe
depression (ADIS-IV-R clinical severity rating >6, see Outcome
Measures), substance abuse or dependence within the last 6
months, psychiatric hospitalization within the last five years,
a history of bipolar disorder, psychosis, mental retardation, or
organic brain damage. If taking psychotropic medication, Ps were
required to be stabilized: one month for benzodiazepines, or three
months for SSRIs/SNRIs/heterocyclics. Please refer to Arch et al. (in
press) for additional study criteria, assessment, and assessment
training procedures.
Measures
Session-by-session measures (purported mediators)
The Anxiety Sensitivity Index (ASI) and a brief version of the
Believability of Anxious Feelings and Thoughts questionnaire
(BAFT) were administered as measures of therapy process at the
beginning of alternating therapy sessions (Sessions 2, 4, 6, 8, 10).
Clients completed these questionnaires in the clinic waiting room
2
Participants who dropped out of treatment after Session 6 (2 by the PI, 5 on
their own) were invited to complete the post-treatment assessments but generally
declined due to time constraints (assessments included a 2e3 h laboratory
assessment in addition to an extensive clinical interview and questionnaires).
3
Mood disorder data was missing for 1 participant.
471
before each treatment session, placed them in a large envelope, and
turned them in to the clinic staff (i.e. not to the study therapists).
The ASI was conceptualized as a CBT process measure whereas the
BAFT was conceptualized as an ACT process measure.
Anxiety Sensitivity Index (ASI). The Anxiety Sensitivity Index
(Peterson & Reiss, 1992; Reiss, Peterson, Gursky, & McNally, 1986)
assesses fear of anxiety-related symptoms (e.g., rapid heart beat)
based on the belief that such sensations have negative social,
physical, or mental consequences. In that fear and negative beliefs
about experiencing fear and anxiety are common features of
anxiety disorders (Craske, Rauch, et al., 2009), the ASI captures
a core component of these disorders. ASI total scores are elevated
across all anxiety disorders (except specific phobia, see Taylor,
Koch, & McNally, 1992) relative to nonclinical samples, and
particularly panic disorder and posttraumatic stress disorder
(Olatunji & Wolitzky-Taylor, 2009). Psychometrically, the ASI has
good internal consistency (a ¼ .82e.91) and stable testeretest
reliability over a three-year period (r ¼ .71) (Maller & Reiss,
1992). ASI assessed at pre- and post-treatment has been shown
to mediate CBT for panic disorder outcomes (Smits et al., 2004).
Current study a’s ranged from .87 to .90 depending on the assessment point.
Believability of Anxious Feelings and Thoughts (BAFT). The original
Believability of Anxious Feelings and Thoughts questionnaire was
a 30-item measure designed to assess cognitive defusion processes
within ACT for anxiety disorders (see Eifert et al., 2009). More
extensive psychometric analyses (BAFT; Herzberg et al., in press)
resulted in a version with 16 items. Higher scores indicate lower
levels of cognitive defusion (or higher levels of cognitive fusion).
Previous ACT studies (e.g., Bach & Hayes, 2002; Gaudiano & Herbert,
2006b; Twohig et al., 2006) have developed similar measures on an
as-needed basis to assess treatment-related change in cognitive
defusion, a central hypothesized ACT process (see Hayes et al.,
2006). In contrast to the non-psychometrically validated cognitive defusion measures used in past research, the BAFT is wellvalidated with high internal consistency in both healthy (a ¼ .90)
and highly anxious (a ¼ .90) samples, and possesses good divergent, convergent, and predictive validity in a treatment-seeking
high anxiety sample (Herzberg et al., in press). We administered
the entire BAFT at pre- and post-treatment, and a briefer 7-item
version as a session-by-session measure to reduce participant
burden. We selected these 7 items from the original BAFT based on
their strong face validity. Six of these 7 items are included in the
revised BAFT (items 1, 4, 8, 14, 15, 16) and demonstrate an overall
BAFT factor loading of M ¼ .65 (range .61e.70). Current study a’s
ranged from .79 to .864 across assessment points.
Treatments
Following the baseline eligibility and diagnostic assessments, Ps
were randomly assigned to CBT or ACT. Ps received twelve weekly,
1-h individual CBT or ACT therapy sessions based on detailed
treatment manuals delivered by doctoral student therapists.5 The
principal authors of the treatment manuals for CBT and ACT led
4
Comparing these 7 BAFT items to the 6 included in the revised BAFT shows that
Cronbach a’s at each time point ranged from .75 to .85 (M ¼ .81) for the 6 item
version and .79 to .86 (M ¼ .82) for the 7-item version, i.e. a’s were identical or
increased when the 7th item is added. Therefore, it seemed statistically sound in
the current sample to include the 7th item despite its exclusion from the revised
BAFT.
5
See author MGC for a copy of the CBT treatment manual; the ACT manual is
published (Eifert & Forsyth, 2005).
472
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
weekly hour-long group supervision for study therapists.6 See Arch
et al. (in press), for additional details of therapist training,
randomization, and treatment.
Traditional cognitive behavioral therapy (CBT). Traditional CBT for
anxiety disorders followed a protocol (Craske, 2005) successfully
employed in previous studies (Craske, Rose, et al., 2009; Craske
et al., 2011) which included general treatment components
shared across the anxiety disorders along with a ‘branching’
mechanism that listed specific cognitive restructuring strategies
and exposure methods for each anxiety disorder. CBT began in
Session 1 with problem assessment, self-monitoring, and psychoeducation. Cognitive restructuring (logical empiricism, evidencebased analysis and hypothesis testing), self-monitoring, and
breathing retraining were introduced in Sessions 2e4. Exposure,
including interoceptive, in-vivo, and imaginal exposure, was
introduced in Session 5 and emphasized strongly in sessions 6e11.
Cognitive restructuring and exposure were tailored to the client’s
principal diagnosis. Clients were taught to answer a series of
questions immediately following each exposure: 1) Did the feared
catastrophe (e.g., collapsing, losing control) come true?; 2) Did I
survive being distressed or afraid?; and 3) Did the fear subside with
repetition? Through this framing, exposure in CBT focused on
hypothesis testing and anxiety reduction over time. Session 12
focused on relapse prevention including planning additional
hypothesis testing exercises and exposures as-needed.
Acceptance and commitment therapy (ACT). ACT for anxiety disorders followed a manual (Eifert & Forsyth, 2005; Eifert et al., 2009);
Session 1 focused on psychoeducation and an orientation to
treatment. Session 2 emphasized “creative hopelessness”, or
exploring whether the client’s efforts to manage and control
anxiety had been successful and experiencing the costs of these
efforts (e.g., diminishment or elimination of valued life activities).
Sessions 3e5 emphasized mindfulness, acceptance, and cognitive
defusion, or the process of distancing oneself from the literal
meaning of anxiety-related thoughts, obsessions, and self-talk.
Acceptance was explored as an alternative to controlling anxiety
via experiential exercises (such as Chinese finger traps, see Eifert &
Forsyth, 2005) and practicing acceptance- and mindfulness-based
meditations in session and at home. Sessions 6e11 continued to
hone acceptance, mindfulness, and cognitive defusion skills and
also emphasized values clarification with the goal of increasing
clients’ willingness to engage in valued life activities on a daily
basis. Personal values were explored via experiential exercises and
Ps were encouraged to behave in ways that reflected their values
rather than spend time managing anxiety. In-vivo and interoceptive exposures were employed during Sessions 6e11 to match the
CBT condition on amount of exposure, and were framed as opportunities to practice acceptance and defusion skills and engage in
valued activities while experiencing anxiety. In the final session
(Session 12), clients discussed how to skillfully manage obstacles
while pursuing a meaningful and value-guided life.
Outcome measures
We assessed whether changes in ASI and BAFT during treatment
mediated outcomes at post-treatment. Given that CBT focused
explicitly on symptom reduction whereas ACT focused on broader
goals of valued living, we investigated both anxiety disorder specific
(e.g., symptom reduction related) and nonspecific or broader
6
For the second half of the study, ACT supervision was led by a highly ACTexperienced doctoral candidate working with Dr. Steven Hayes at the University
of Nevada at Reno.
outcomes across both treatments. Clinical severity ratings for the
principal anxiety disorder diagnosis, clinically relevant worry, and
anxiety-related behavioral avoidance comprised the anxiety disorder
specific outcomes. Worry and behavioral avoidance were selected
because they represent central shared features of anxiety disorders
(Craske, Rauch, et al., 2009). Quality of life and depression symptoms7
comprised the nonspecific, broader outcomes. The questionnaires
assessing the outcome indices are described briefly. At pre-treatment,
CBT and ACT did not differ on any outcomes, ps .10.
Anxiety disorder specific outcomes
Diagnostic interview assessment. The information extracted from
the ADIS-IV-R included ‘clinical severity ratings’ (CSR; distress and
disability) made by group consensus on a 0 to 8 scale (0 ¼ none,
8 ¼ extreme) for each disorder. Baseline ratings of “4” or higher
indicated a clinically significant degree of distress or disability;
ratings of 4 or higher for at least one anxiety disorder served as the
cutoff for study eligibility (see Craske, DeCola, Sachs, & Pontillo,
2003; Craske et al., 2007). Averaging across the various principal
anxiety disorders, the CSR pre-treatment mean was 5.7 (SD ¼ .93,
range 4e8) and post-treatment mean was 3.0 (SD ¼ 2.2,
range ¼ 0e7). Pre to post reductions in principal anxiety disorder
CSR ratings did not differ by treatment group (see Arch et al., in
press, for more detailed analyses).
To assess inter-rater reliability, ADIS-IV-R assessments were
audiotaped and 15% of Ps in the parent study (n ¼ 22) were selected
at random for blind ratings by a second rater. One-way intraclass
correlation coefficients for diagnostic status (clinically significant
distress and disability [CSR of 4þ] vs. subclinical [CSR of 1e3] vs.
none [CSR of 0]) were social anxiety disorder and obsessive
compulsive disorder ICC ¼ 1.00 (100% agreement), panic disorder
ICC ¼ .91, generalized anxiety disorder ICC ¼ .85, and specific
phobia ICC ¼ .75. Dimensional CSRs across all principal diagnoses
evidenced an ICC of .658 (Arch et al., in press).
The Penn State Worry Questionnaire (PSWQ; Meyer, Miller, Metzger,
& Borkovec, 1990) is a 16-item measure of trait worry which focuses
on the intensity, excessiveness, and uncontrollability of clinically
relevant worry (Molina & Borkovec, 1994). The PSWQ demonstrates
good internal consistency (a of .86 to .93 across clinical and college
samples) and testeretest reliability (r ¼ .74e.93 across 2e10 week
periods). Although particularly elevated in GAD, the PSWQ shows
elevations across all anxiety disorders relative to non-anxious controls
(Brown, Antony, & Barlow,1992). The current sample demonstrated as
of .89 (pre-treatment) and .93 (post-treatment).
The Fear Questionnaire (FQ; Marks & Mathews, 1979) is a questionnaire that assesses fear, anxiety, and behavioral avoidance
among different anxiety disorders. The Main Target Phobia Scale, an
avoidance rating for each P’s “main phobia”, was used as the
behavioral avoidance outcome.
Broader outcomes
The Quality of Life Inventory (Frisch, 1994) is a 32-item questionnaire assessing life satisfaction across 16 broad life domains.
7
We also assessed experiential avoidance/acceptance with the 16-item Acceptance and Action Questionnaire (AAQ; Bond & Bunce, 2000; Hayes et al., 2004) and
considered using it as a broader treatment outcome. At post-treatment, however,
the AAQ and full BAFT correlated at r ¼ .71, indicating a high degree of overlap
between these measures. Due to this overlap, we did not use the AAQ as an
outcome measure in the mediator analyses. For analyses of AAQ outcomes, see Arch
et al. (in press).
8
The ICC across all principal diagnoses was lower than for individual diagnoses
because the individual diagnoses included no symptoms (e.g., none) and were rated
on a simpler 3-point scale (relative to the 0e8 CSR scale), which promoted a higher
level of agreement.
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
473
Quality of life is calculated by multiplying an importance rating by
satisfaction rating for each area and summing them to produce an
overall quality of life score. Psychometric properties include good
testeretest reliability (at 2-week intervals r ¼ .73), as well as good
convergent, discriminant, and treatment validity (Frisch et al.,
2005). The as in the current sample were .83 (pre-treatment) and
.85 (post-treatment).
The General Depression scale, Mood and Anxiety Symptom Questionnaire (Watson & Clark, 1991) is a 12 item measure of depression
symptoms that demonstrates good divergent (r ¼ .02e.09) and
convergent validity (r ¼ .67e.76) with other depression scales and
strong incremental validity across student, adult, and patient
samples (Watson, Clark, et al., 1995; Watson, Weber, et al., 1995).
Current as were .88 (pre-treatment) and .94 (post-treatment).
Results
Analyses were performed in HLM 6.0 (Raudenbush, Bryk,
Cheong, & Congdon, 2004) and SPSS 16.0. Analyses included Ps
with post-treatment data as well as at least one session-by-session
data point. Thus, Ps with missing data were included in the analyses
as long as they met these criteria.9 ASI and BAFT data points refer to
alternating-session scores on the ASI and BAFT (7-item version)
from baseline through Session 10 (6 total data points). We
computed effect sizes (d) based on the magnitude effect size
formula for HLM analyses presented by Feingold (2009) based on
earlier work by Raudenbush and Liu (2001). Statistical significance
was defined as p < .05.
We assessed whether the presence or amount of missing data
predicted any of the post-treatment outcomes that were linked to
the mediators (covarying outcome measures at pre-treatment). On
average, 32.4% of data was missing at each session-by-session time
point; the majority of Ps (83.6%) had 2 or fewer sessions of missing
data. Importantly, the presence of missing data did not predict any
of the post-treatment outcomes (all ps > .13). The amount of
missing data for each participant also did not predict behavioral
avoidance, depression, or quality of life outcomes (ps > .11), though
approached significance for worry outcomes (p ¼ .06). In summary,
there were no statistically significant associations between missing
data and the predicted outcomes.
Decline in ASI and BAFT across assessment periods during treatment
A series of 2-level hierarchical linear growth curve models
(HLMs) were conducted with ASI or BAFT during treatment as the
outcome variable, Time as the level-1 predictor, and Group (CBT vs.
ACT) as the level-2 predictor. In the first model, baseline ASI scores
(i.e. intercept) were borderline significantly higher in ACT,
b ¼ 33.59, than in CBT, b ¼ 28.48, between group b ¼ 5.12, t
(65) ¼ 1.96, p ¼ .053, d ¼ .47. ASI significantly declined across time
in CBT, b ¼ 2.16, t (65) ¼ 5.46, p < .001, d ¼ 1.0010, and in ACT,
b ¼ 3.70, t (65) ¼ 7.34, p < .001, d ¼ 1.71 with those in ACT
1.55, t
showing steeper ASI decline slopes than CBT, b ¼
(65) ¼ 2.80, p ¼ .007, d ¼ .71 (see Fig. 1a).
In the second model, baseline BAFT scores did not differ
between CBT, b ¼ 18.75, and ACT, b ¼ 18.45, between group b ¼ .30,
9
Ps who were missing all mediator and post-treatment data, i.e. those who did
not begin treatment or who were treated prior to the implementation of the
mediator measures, were logically excluded because it did not make sense to
examine treatment mediators for Ps who lacked treatment data.
10
Effect sizes characterize change during treatment, that is, from baseline to
Session 10 (of 12 Sessions), and thus exclude post-treatment data. Assuming
a constant linear change rate, they are approximately 20% smaller than if posttreatment data were included.
Fig. 1. a. Change in anxiety sensitivity during treatment. b. Change in cognitive fusion
during treatment.
t (65) ¼ .33, p ¼ .74, d ¼ .08. BAFT significantly declined across
time11 in CBT, b ¼ 1.36, t (65) ¼ 6.39, p < .001, d ¼ 1.85, and in
ACT, b ¼ 1.96, t (65) ¼ 9.27, p < .001, d ¼ 2.67, with ACT
showing nearly significantly steeper BAFT decline slopes than CBT,
b ¼ .60, t (65) ¼ 1.99, p ¼ .050, d ¼ .82 (see Fig. 1b).
BAFT and ASI as mediators of treatment outcome
Generalized least squares estimates analysis was performed
within HLM to obtain y-intercept (i.e. baseline score on ASI or BAFT)
and decline slope (i.e. change over time on ASI or BAFT during
treatment) parameter estimates for each participant. These slope
and intercept parameters were then entered as predictors into
a series of linear regression analyses (described below). We used
the MacArthur guidelines as outlined by Kraemer et al. (2002) to
test for mediation. In this approach, a variable can be considered
a mediator of treatment outcome if: (a) it is measured during
treatment; (b) rate of change during treatment is correlated with
treatment Group; and (c) it either has a direct relation with the
outcome variable or interacts with the treatment Group in its
relation to the outcome.
For all analyses including BAFT and ASI slopes as putative
mediators, criterion (a) was clearly met, as these variables were
measured during treatment. Criterion (b) was also met since
change in BAFT and ASI during treatment correlated with Group
(see Results above; albeit only a borderline significant [p ¼ .050]
group effect for the BAFT). Therefore, for each mediation analysis
below, we focus on whether criterion (c) was met. If criterion (c) is
11
Please note that declines in BAFT indicate increases in cognitive defusion (i.e.
decreases in cognitive fusion).
474
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
met for the ASI analyses below, the requirements for treatment
mediation were met fully; if criterion (c) is met for the BAFT analyses, the requirements for treatment mediation were met largely
but our conclusions are more cautious given the partial fulfillment
of criterion (b).
Regression analyses in SPSS were conducted for the y-intercept
with the putative mediating variable (i.e. estimated baseline score
on BAFT or ASI) in the first block to adjust for baseline scores on
these measures, Group and BAFT or ASI slope (i.e. main effect
terms) in the second block, and their interaction term (BAFT or ASI
slope Group) in the third block. That is, the Group Time
interaction effect was tested on the final block of the regression
equation. Significant Group Time interaction effects indicated
mediated moderation, that is, the mediation of treatment outcome
that was moderated by treatment group. If the Group Time
interaction term was non-significant, the analysis was re-run
without it to examine the main effect of the mediator. Significant
main effects indicated that mediation took place but was not
moderated by treatment group (e.g., no moderated mediation).
The dependent variable was the outcome variable at posttreatment (see Outcome Measures). Slopes represented change
in the putative mediators during treatment (post-treatment scores
were excluded from slope estimates) and therefore, temporally
preceded the post-treatment outcomes that served as the dependent variable. By measuring the putative mediators prior to the
post-treatment outcomes, we have established their temporal
precedence. Separate analyses were conducted for BAFT and ASI
mediation. Note that the reason we did not conduct the entire
analysis in HLM is because our outcome variables were single data
points (e.g., the post-treatment outcomes) whereas in HLM,
outcome variables must be time varying or longitudinal. See
Table 1 for significant regression results, which are summarized
below.
Anxiety disorder specific outcomes
Does change on the BAFT and/or ASI mediate clinical severity
ratings for the principal diagnosis? Within the model described
above, BAFT slope Group interaction did not significantly predict
principal diagnosis CSR at post-treatment (p ¼ .72). A simplified
model removing the interaction term was conducted. Still there
was no significant main effect of BAFT slope on CSR at posttreatment (p ¼ .21). Similarly, in the final model described above,
the ASI Group interaction did not significantly predict principal
diagnosis CSR at post-treatment (p ¼ .21); nor did a simplified
model removing the interaction term, ASI slope (p ¼ .47). Thus,
neither putative mediator was found to mediate the primary
outcome measure of clinical severity.
Does change on the BAFT and/or ASI mediate change in worry
symptoms? A significant interaction of BAFT slope Group was
8.98,
observed in predicting post-treatment PSWQ, b ¼
t (39) ¼ 2.25, p < .05, Adjusted R2 ¼ .44, DR2 ¼ .07. Thus, criterion
(c) was met and BAFT served as a mediator of change in worry
outcomes. The nature of this interaction was observed by plotting
estimated regression lines, which showed that overall, BAFT decline
was associated with lower PSWQ scores, and that those in CBT with
steeper BAFT decline slopes had lower PSWQ scores than those in
ACT with similarly steep decline slopes (see Fig. 2).
An identical model was run with ASI variables. The ASI
slope Group interaction was not significant (p ¼ .89). A simplified
model removing the interaction term was conducted to examine
main effects. ASI slope significantly predicted PSWQ scores at posttreatment, b ¼ 11.73, t (39) ¼ 2.95, p < .01, Adjusted R2 ¼ .34,
DR2 ¼ .15, such that steeper ASI decline slopes over treatment were
associated with lower PSWQ scores at post-treatment across both
groups. This result fulfilled the criterion (c) requirement
Table 1
Linear regression statistics for mediation analyses.
B
BAFT as a mediator of PSWQ outcome
Model 1 (interaction)
Pre-treatment PSWQ
.30
BAFT intercept
3.50
BAFT slope
15.26
Group
8.11
BAFT slope Group
8.99
ASI as a mediator of PSWQ outcome
Model 1 (interaction)
Pre-treatment PSWQ
.47
ASI intercept
2.64
ASI slope
12.01
Group
6.20
ASI slope Group
.41
Model 2 (main effects)
Pre-treatment PSWQ
.47
ASI intercept
2.67
ASI slope
11.73
Group
6.11
BAFT as a mediator of FQ1 outcome
Model 1 (interaction)
Pre-treatment FQ1
.001
BAFT intercept
1.01
BAFT slope
.87
Group
1.14
BAFT slope Group
1.27
Model 2 (main effects)
Pre-treatment FQ1
.001
BAFT intercept
.99
BAFT slope
1.46
Group
1.08
ASI as a mediator of FQ1 outcome
Model 1 (interaction)
Pre-treatment FQ1
.001
ASI intercept
.15
ASI slope
1.25
Group
1.09
ASI slope Group
.73
Model 2 (main effects)
.001
Pre-treatment FQ1
ASI intercept
.14
ASI slope
1.73
Group
1.09
BAFT as a mediator of QOLI outcome
Model 1 (interaction)
Pre-treatment QOLI
.66
BAFT intercept
12.73
BAFT slope
10.87
Group
3.82
BAFT slope Group
2.54
Model 2 (main effects)
Pre-treatment QOLI
.65
BAFT intercept
13.00
BAFT slope
12.51
Group
4.04
SE
t
p
.16
1.99
3.79
3.90
4.00
1.87
1.75
4.03
2.08
2.24
<.01
.09
<.001
<.05
.03
.16
2.47
4.54
4.40
3.04
2.88
1.07
2.64
1.41
.13
<.01
.29
.01
.17
.89
.16
2.42
3.98
4.28
2.92
1.10
2.95
1.43
<.01
.28
<.01
.16
.003
.59
.84
1.00
1.01
.28
1.71
1.04
1.13
1.25
.78
.10
.31
.27
.22
.003
.60
.70
1.01
.44
1.66
2.10
1.06
.66
.11
.04
.30
.003
.57
1.14
1.04
.87
.36
.27
1.10
1.05
.83
.73
.79
.28
.30
.41
.003
.57
.99
1.03
.45
.25
1.75
1.05
.66
.81
.09
.30
.10
3.34
5.86
5.75
6.78
6.30
3.81
1.86
.66
.38
<.001
.001
.07
.51
.71
.10
3.23
3.84
5.65
6.44
4.03
3.26
.71
<.001
<.001
<.01
.48
5.79
1.74
1.18
.75
.63
<.001
.09
.25
.46
.53
5.79
1.75
2.17
.83
<.001
<.001
<.05
.41
BAFT as a mediator of MASQ general depression outcome
Model 1 (interaction)
Pre-treatment MASQ
.85
.67
BAFT intercept
3.14
.26
BAFT slope
3.26
.24
Group
2.42
.09
BAFT slope Group
2.00
.11
Model 2 (main effects)
Pre-treatment MASQ
.83
.14
BAFT intercept
3.14
1.79
BAFT slope
4.43
2.04
Group
2.65
3.19
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
475
Discussion
Fig. 2. Cognitive defusion group interaction in predicting post-treatment worry.
for mediation and ASI was considered a mediator of worry
outcomes.
Does change in BAFT and/or ASI mediate changes in behavioral
avoidance? There was no significant BAFT slope Group interaction
on FQ1 (p ¼ .22). The simplified model revealed a significant main
effect of BAFT slope on FQ1, b ¼ 1.46, t (37) ¼ 2.10, p < .05, Adjusted
R2 ¼ .03, DR2 ¼ .13,12 such that steeper BAFT decline slopes were
associated with lower FQ1 at post-treatment. This result fulfilled
the criterion (c) requirement for mediation and BAFT was considered a mediator of behavioral avoidance outcomes. In contrast,
there was neither an ASI slope Group interaction on FQ1 (p ¼ .41)
nor a significant main effect of ASI slope on FQ1 (p ¼ .09). However,
the main effect of ASI slope on FQ1 was in the same direction as that
of the BAFT slope, albeit not statistically significant, Adjusted
R2 ¼ .02, DR2 ¼ .09.
Broader outcomes
Does change on the BAFT and/or ASI mediate change in quality of
life? The BAFT slope Group interaction did not significantly
predict quality of life at post-treatment (p ¼ .71). After removing
the interaction term, there was a significant main effect of BAFT
slope, b ¼ 12.52, t (41) ¼ 3.56, p < .01, Adjusted R2 ¼ .61, DR2 ¼ .10,
such that steeper BAFT declines were associated with increases in
quality of life. This result fulfilled the criterion (c) requirement for
mediation and BAFT was considered a mediator of quality of life
outcomes. The ASI Group interaction did not significantly predict
quality of life (p ¼ .42). In the simplified model, ASI slope did not
significantly predict quality of life at post-treatment (p ¼ .07) but
the effect was in the same direction as that of the BAFT slope,
Adjusted R2 ¼ .48, DR2 ¼ .05.
Does change on the ASI and/or BAFT mediate Depression
outcomes? The BAFT slope Group interaction did not significantly
predict MASQ General Depression (p ¼ .53). In the simplified
model, the BAFT slope significantly predicted depression scores at
post-treatment, b ¼ 4.43, t (42) ¼ 2.17, p < .05, DR2 ¼ .06, such that
steeper BAFT decline slopes were associated with lower depression
at post-treatment. This result fulfilled the criterion (c) requirement
for mediation and BAFT was considered a mediator of depression
outcome.
The models including ASI variables yielded no significant
interaction or main effect terms for MASQ General Depression
scores (ps > .43). Thus, ASI did not mediate depression outcomes.
12
Although the Adjusted R2 is smaller than the DR2, the Total Unadjusted R2 ¼ .13,
which is equivalent or larger than the DR2 for both ASI and BAFT models.
Our analyses addressed two central research questions
regarding CBT or ACT for the treatment of a heterogeneous anxiety
disorder sample. First, we addressed whether CBT and ACT affected
the purported mediators for each treatment (anxiety sensitivity in
CBT, cognitive defusion in ACT). Both mediators evidenced large
effect size improvements during both treatments. ACT resulted in
greater improvements than CBT, of medium to large effect size, for
both anxiety sensitivity and cognitive defusion. Group differences
in cognitive defusion were not fully significant (p ¼ .05), but were
large in magnitude. The finding therefore supported our hypothesis
that cognitive defusion would decline significantly and to a greater
extent in ACT than CBT. Furthermore, the lack of group differences
in cognitive defusion at pre-treatment clarifies that the observed
effects were due to treatment. Our hypothesis that anxiety sensitivity would decline significantly and to a greater extent in CBT than
ACT was not supported, however, since anxiety sensitivity declined
more in ACT. At pre-treatment, ACT showed marginally higher
anxiety sensitivity than CBT, suggesting that ACT’s steeper
improvement in anxiety sensitivity during treatment was due in
part to greater room for improvement. Our second question
addressed whether session-by-session changes in anxiety sensitivity and cognitive defusion mediated treatment outcomes and
whether this differed by group. We categorized outcomes as
anxiety disorder specific vs. broader treatment outcomes. Anxiety
disorder specific outcomes included the clinical severity rating of
the principal diagnosis, clinical worry, and anxiety disorder-related
behavioral avoidance. Broader outcomes included quality of life and
depression.
For anxiety disorder specific outcomes, neither session-bysession change in cognitive defusion nor anxiety sensitivity predicted the clinical severity ratings of the principal diagnosis at posttreatment. Their high variability at post-treatment may account for
the non-significant findings on this outcome. Alternatively,
processes other than the specified mediators, such as treatment
credibility, homework adherence, or therapist adherence,13 may
account for changes in clinical severity. On the other hand, sessionby-session change in anxiety sensitivity and cognitive defusion
both mediated clinical worry outcomes. For anxiety sensitivity,
steeper declines in anxiety sensitivity during treatment predicted
lower worry at post-treatment across both treatment groups, with
a large effect size. For cognitive defusion, a significant interaction of
medium effect size demonstrated that increases in cognitive
defusion during treatment predicted diminished worry at posttreatment to a greater extent in CBT than ACT. This finding was
not due to steeper overall cognitive defusion increases in CBT. In
fact, as noted above, the opposite is true: ACT evidenced steeper
increases in cognitive defusion than CBT. This unexpected finding
suggests that the ACT-specific mediator (cognitive defusion) played
a more central role in predicting worry reduction in the non-ACT
treatment (CBT). We return to this point below. Finally,
increases in cognitive defusion during treatment significantly
mediated reductions in anxiety-related behavioral avoidance at
post-treatment across both treatments, by a medium to large
effect size. Anxiety sensitivity did not significantly mediate
behavioral avoidance outcomes, though outcomes were in the
predicted direction with a smaller effect size than for cognitive
defusion.
13
Per Arch et al. (in press), general therapist competency ratings (i.e. general
therapy skills, non-condition specific) were not related to post-treatment clinical
severity ratings. Other potential predictors of clinical severity ratings will be tested
and reported elsewhere.
476
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
For the broader outcomes, session-by-session change in cognitive defusion significantly mediated both quality of life and
depression outcomes, whereas anxiety sensitivity was not a statistically significant mediator. For quality of life, increases in cognitive
defusion mediated higher post-treatment quality of life across both
treatments with a medium to large effect size. Although not statistically significant, anxiety sensitivity evidenced a borderline significant mediator effect across treatments of medium effect size. The
finding that changing the relationship with anxiety-related
thoughts (via cognitive defusion) was a more robust mediator of
quality of life increases than changing the content of anxiety and
fear-related beliefs (via anxiety sensitivity) is consistent with the
ACT model. According to the ACT model of psychopathology (Hayes
et al., 2006, 2012), increasing cognitive defusion results in greater
psychological flexibility, or the capacity to respond flexibly according to all available options and move forward in valued life directions
rather than react rigidly to the content of (anxiety-related) thoughts
and other unhelpful conceptions. Psychological flexibility creates
space to enact personally valued actions (rather than anxiety-driven
actions), which should lead to higher quality of life. Further, by
decreasing the narrow, rigid sense that life is dictated by the need to
respond to anxiety-related cognitions, cognitive defusion should
decrease despondency, avolition, and associated depressive symptoms. In support of this notion, increases in cognitive defusion predicted lower depression at post-treatment across both treatment
groups, with a medium effect size. Cognitive defusion in ACT
parallels the construct of decentering in Mindfulness Based Cognitive Therapy, in which thoughts and feelings are viewed from
a detached perspective as ‘mental events’ that are not necessarily
aspects of the self or accurate reflections of reality (Segal, Williams, &
Teasdale, 2002). It is notable that even in an anxiety disorder sample,
defusion/decentering mediated improvements in depression.
In summary, cognitive defusion e the capacity to not be rigidly
governed by cognitive content e appears to be an important
mediator of both CBT and ACT for anxiety disorders. Cognitive
defusion significantly mediated all post-treatment outcomes
except one (4/5 outcomes) whereas anxiety sensitivity significantly
mediated only one outcome, although two others evidenced
borderline significant effects. Across both ACT and CBT, therefore,
cognitive defusion served as a stronger mediator of post-treatment
outcomes than anxiety sensitivity. Further, in CBT, cognitive defusion more strongly mediated worry outcomes than in ACT, suggesting that even within the CBT model that does not explicitly
discuss cognitive defusion, this process occurs, and underlies
change to the same or greater extent than in ACT. Several CBT
strategies implicitly promote cognitive defusion, including selfmonitoring and cognitive restructuring, both of which aim to
cultivate a ‘scientific observer perspective’ and therapeutic distance
from anxiety-related cognitions. In ACT, cognitive defusion-based
exercises more directly target this process (see Arch & Craske,
2008). Not surprisingly, given its more direct targeting in ACT,
cognitive defusion increased more during ACT than CBT. Despite
the different approaches to increasing cognitive defusion in each
CBT and ACT, improvement on this variable represents a key therapeutic process across both therapies.
The overall findings do not provide evidence that CBT and ACT
for anxiety disorders differ substantially in terms of mediation
pathways, at least with regard to anxiety sensitivity and cognitive
defusion. We have previously theorized that CBT and ACT for
anxiety disorders share some common underlying therapeutic
processes or mediators (Arch & Craske, 2008). To our knowledge,
this is the first empirical test of this hypothesis.
Several study limitations should be noted. First, our outcome
measures were limited to independent diagnostic evaluations and
self-report questionnaires and our mediators were limited to self-
report questionnaires only. Because questionnaires rely on oftenlimited conscious awareness of mental processes, future studies
would benefit from integrating behavioral, psychophysiological, and
brain-based measurements of mediators and outcomes. Second, we
could not test more than one hypothesized mediator of each treatment model due to limited patient time (patients had to complete
mediators immediately prior to alternating therapy sessions, and
often ran somewhat late for sessions). Future studies should aim to
assess a variety of different potential mediators, including thought
frequency, controllability, and/or content for CBT, using both implicit
and explicit paradigms, and self-as-context (e.g., self-defusion),
experiential avoidance, acceptance, behavioral commitments, or
values clarification for ACT. Third, ACT and CBT were matched on the
number of sessions devoted to behavioral exposure, which likely
minimized group differences in mediation pathways. The original
ACT manual (Hayes et al., 1999) did not emphasize traditional
exposure methods. Fourth, for the ACT mediator we used a brief
version of a little-known measure (BAFT) relative to the widely used
Anxiety Sensitivity Index for CBT. Well-validated measures of ACT
processes, however, were still under development when we began
this study. The BAFT represents one of the best validated among
them and is the only validated ACT process measure for disordered
anxiety. In addition, although anxiety sensitivity is elevated among
all of the anxiety disorders assessed in this study (see Taylor et al.,
1992), it is higher in panic disorder and posttraumatic stress
disorder (Olatunji & Wolitzky-Taylor, 2009) and therefore may
represent a more robust mediator within those disorders. It is challenging, however, to find a psychometrically sound CBT-relevant
mediator with strictly equal relevance across all of the anxiety
disorders. As such, anxiety sensitivity may be among the best
candidates. Fifth, our use of beginning therapists who were relatively
inexperienced in CBT and new to ACT may have reduced the overall
effectiveness of both treatments. Sixth, our mixed anxiety disorder
sample limits the conclusions that can be drawn regarding mediation for a specific anxiety disorder. Future studies will need to assess
whether our findings hold for individual anxiety disorders. On the
other hand, given the overlapping features of anxiety disorders
(Craske, Rauch, et al., 2009; Watson, 2005) a strength of this
approach was that it facilitated an examination of mediation across
a broader nosologic category. Seventh, although we established
temporal precedence for our mediators, they were measured
through session 10 and a session 10 mediator is likely to be correlated
with the outcomes assessed (several weeks) after the final session
(session 12). It can be challenging, therefore, to narrow precisely
when one variable influenced another. Nonetheless, this study
represents a preliminary step in identifying treatment mediators and
progresses beyond most previous studies with regard to temporal
precedence. Finally, this study was underpowered to detect between
group differences that were small in magnitude. Future studies
should aim to replicate the present findings in larger samples.
In summary, this study, to our knowledge, represents the first
empirical investigation of treatment mediation in CBT vs. ACT for
anxiety disorders. Cognitive defusion emerged as a robust mediator
of outcomes within both CBT and ACT; anxiety sensitivity mediated
outcomes to a less significant degree. Findings are consistent with
our previous assertion that treatment mediation is partially shared
across both CBT and ACT for anxiety disorders (Arch & Craske,
2008). Further, findings suggest that cognitive defusion contributes considerably to the success of behaviorally-based anxiety
disorder treatments.
Acknowledgments
First author J.J.A. was assisted in the preparation of this manuscript by startup funds from the University of Colorado Boulder.
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
References
Arch, J. J., & Craske, M. G. (2008). Acceptance and commitment therapy and
cognitive behavioral therapy for anxiety disorders: different treatments, similar
mechanisms? Clinical Psychology: Science and Practice, 15(4), 263e279.
Arch, J. J., Eifert, G. H., Davies, C., Plumb, J. C., Rose, R. D., & Craske, M. G. Randomized
clinical trial of cognitive behavioral therapy versus acceptance and
commitment therapy for the treatment of mixed anxiety disorders. Journal of
Consulting and Clinical Psychology, in press.
Bach, P., & Hayes, S. C. (2002). The use of acceptance and commitment therapy to
prevent the rehospitalization of psychotic patients: a randomized controlled
trial. Journal of Consulting and Clinical Psychology, 70(5), 1129e1139.
Bond, F. W., & Bunce, D. (2000). Mediators of change in emotion-focused and
problem-focused worksite stress management interventions. Journal of Occupational Health Psychology, 5(1), 156e163.
Brown, T. A., Antony, M. M., & Barlow, D. H. (1992). Psychometric properties of the
Penn State Worry Questionnaire in a clinical anxiety disorders sample.
Behaviour Research and Therapy, 30(1), 33e37.
Butler, A. C., Chapman, J. E., Forman, E. M., & Beck, A. T. (2006). The empirical status
of cognitive-behavioral therapy: a review of meta-analyses. Clinical Psychology
Review, 26, 17e31.
Craske, M. G. (2005). Cognitive-behavioral treatment of anxiety disorders.
Unpublished manuscript.
Craske, M. G. (2010). Cognitive-behavioral therapy. Washington, D.C.: American
Psychological Association.
Craske, M. G., DeCola, J. P., Sachs, A. D., & Pontillo, D. C. (2003). Panic control
treatment for agoraphobia. Journal of Anxiety Disorders, 17(3), 321e333.
Craske, M. G., Farchione, T. J., Allen, L. B., Barrios, V., Stoyanova, M., & Rose, R. (2007).
Cognitive behavioral therapy for panic disorder and comorbidity: more of the
same or less of more? Behaviour Research and Therapy, 45(6), 1095e1109.
Craske, M. G., Rauch, S. L., Ursano, R., Prenoveau, J., Pine, D. S., & Zinbarg, R. E.
(2009). What is an anxiety disorder? Depression and Anxiety, 26, 1066e1085.
Craske, M. G., Rose, R. D., Lang, A., Welch, S. S., Campbell-Sills, L., Sullivan, G., et al.
(2009). Computer-assisted delivery of cognitive behavioral therapy for anxiety
disorders in primary-care settings. Depression and Anxiety, 26(3), 235e242.
Craske, M. G., Stein, M. B., Sullivan, G., Sherbourne, C., Bystritsky, A., Rose, R. D., et al.
(2011). Disorder-specific impact of coordinated anxiety learning and
management treatment for anxiety disorders in primary care. Archives of
General Psychiatry, 68(4), 378e387. doi:10.1001/archgenpsychiatry.2011.25.
Dalrymple, K., & Herbert, J. D. (2007). Acceptance and commitment therapy for
generalized social anxiety disorder: a pilot study. Behavior Modification, 31(5),
543e568.
DiNardo, P. A., Brown, T. A., & Barlow, D. H. (1994). Anxiety Disorders Interview
Schedule for DSM-IV (ADIS-IV). Albany, NY: Graywind Publications.
Eifert, G. H., & Forsyth, J. P. (2005). Acceptance and commitment therapy for anxiety
disorders: A practitioner’s treatment guide to using mindfulness, acceptance, and
values-based behavior change strategies. New York, NY: Guilford Press.
Eifert, G. H., Forsyth, J. P., Arch, J. J., Keller, M., Langer, D., & Espejo, N. (2009).
Acceptance and commitment therapy for anxiety disorders: case studies using
a unified treatment protocol. Cognitive and Behavioral Practice, 16, 368e385.
Feingold, A. (2009). Effect sizes for growth-modeling analysis for controlled clinical trials
in the same metric as for classical analysis. Psychological Methods, 14(1), 43e53.
Forman, E. M., Herbert, J. D., Moitra, E., Yeomans, P. D., & Geller, P. A. (2007).
A randomized controlled effectiveness trial of acceptance and commitment
therapy and cognitive therapy for anxiety and depression. Behavior
Modification, 31(6), 772e799.
Forsyth, J. P., Eifert, G. H., & Barrios, V. (2006). Fear conditioning in an emotion
regulation context: a fresh perspective on the origins of anxiety disorders. In
M. G. Craske, D. Hermans, & D. Vansteenwegen (Eds.), From basic processes to
clinical implications. Washington, D.C.: American Psychological Association.
Frisch, M. B. (1994). Quality of Life Inventory (QOLI). Minneapolis: National Computer
Systems.
Frisch, M. B., Clark, M. P., Rouse, S. V., Rudd, M. D., Paweleck, J., & Greenstone, A.
(2005). Predictive and treatment validity of life satisfaction and the Quality of
Life Inventory. Assessment, 12, 66e78.
Gaudiano, B. A., & Herbert, J. D. (2006a). Acute treatment of inpatients with
psychotic symptoms using acceptance and commitment therapy: Pilot results
[Peer Reviewed]. Behaviour Research and Therapy, 44(3), 415e437. doi:10.1016/
j.brat.2005.02.007.
Gaudiano, B. A., & Herbert, J. D. (2006b). Believability of hallucinations as a potential
mediator of their frequency and associated distress in psychotic inpatients [Peer
Reviewed]. Behavioural and Cognitive Psychotherapy, 34(4), 497e502.
doi:10.1017/s1352465806003080.
Hayes, S. C., Luoma, J. B., Bond, F. W., Masuda, A., & Lillis, J. (2006). Acceptance and
commitment therapy: model, processes and outcomes. Behaviour Research and
Therapy, 44(1), 1e25.
Hayes, S. C., Strosahl, K. D., & Wilson, K. G. (1999). Acceptance and commitment therapy:
An experiential approach to behavior change. New York, NY: Guilford Press.
Hayes, S. C., Strosahl, K. D., & Wilson, K. G. (2012). Acceptance and commitment
therapy: The process and practice of mindful change (2nd ed.). New York, NY:
Guilford Press.
Hayes, S. C., Strosahl, K., Wilson, K. G., Bissett, R. T., Pistorello, J., Toarmino, D., et al.
(2004). Measuring experiential avoidance: a preliminary test of a working
model. Psychological Record, 54(4), 553e578.
477
Herzberg, K. N., Sheppard, S. C., Forsyth, J. P., Eifert, G. H., Crede, M., & Earleywine,
M. The believability of anxious feelings and thoughts questionnaire (BAFT):
psychometric properties in a healthy undergraduate and high-anxious
community sample. Psychological Assessment, in press.
Hofmann, S. G. (2004). Cognitive mediation of treatment change in social phobia.
Journal of Consulting and Clinical Psychology, 72(3), 392e399.
Hofmann, S. G., & Smits, J. A. J. (2008). Cognitive-behavioral therapy for adult
anxiety disorders: a meta-analysis of randomized placebo-controlled trials.
Journal of Clinical Psychiatry, 69(4), 621e632.
Hofmann, S. G., Suvak, M. K., Barlow, D. H., Shear, M. K., Meuret, A. E., Rosenfeld, D.,
et al. (2007). Preliminary evidence for cognitive mediation during
cognitive-behavioral therapy of panic disorder. Journal of Consulting and Clinical
Psychology, 75(3), 374e379.
Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In
(4th ed.).. In Gilbert, D. T., Fiske, S. T., & Lindzey, G. (Eds.), The handbook of
social psychology, vols. 1 and 2 (pp. 233e265) New York, NY, US: McGraw-Hill, x,
1085.
Kessler, R. C., Berglund, P., Demler, O., Jin, R., & Walters, E. E. (2005). Lifetime
prevalence and age-of-onset distributions of DSM-IV disorders in the
national comorbidity survey replication. Archives of General Psychiatry, 62(6),
593e602.
Kraemer, H. C., Wilson, T., Fairburn, C. G., & Agras, W. S. (2002). Mediators and
moderators of treatment effects in randomized clinical trials. Archives of General
Psychiatry, 59(10), 877e883.
Lappalainen, R., Lehtonen, T., Skarp, E., Taubert, E., Ojanen, M., & Hayes, S. C. (2007).
The impact of CBT and ACT models using psychology trainee therapists:
a preliminary controlled effectiveness trial. Behavior Modification, 31(4),
488e511.
MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual
Review of Psychology, 58, 593e614.
Maller, R. G., & Reiss, S. (1992). Anxiety sensitivity and panic attacks: a longitudinal
analysis. Journal of Anxiety Disorders, 6, 241e247.
Marks, I. M., & Mathews, A. M. (1979). Brief standard self-rating for phobic patients.
Behaviour Research and Therapy, 17, 263e267.
Meuret, A. E., Rosenfield, D., Hofmann, S. G., Seidel, A., & Bhaskara, L. (2010).
Respiratory and cognitive mediators of treatment change in panic disorder:
evidence for intervention specificity. Journal of Consulting and Clinical
Psychology, 78(5), 691e704.
Meuret, A. E., Rosenfield, D., Hofmann, S. G., Suvak, M. K., & Roth, W. T. (2009).
Changes in respiration mediate changes in fear of bodily sensations in panic
disorder. Journal of Psychiatric Research, 43, 634e641.
Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D. (1990). Development and
validation of the Penn State Worry Questionnaire. Behaviour Research and
Therapy, 28, 487e495.
Molina, S., & Borkovec, T. D. (Eds.). (1994). The Penn State Worry Questionnaire:
Psychometric properties and associated characteristics. Oxford, England: John
Wiley & Sons.
Moscovitch, D. A., Hofmann, S. G., Suvak, M. K., & In-Albon, T. (2005). Mediation of
changes in anxiety and depression during treatment of social phobia. Journal of
Consulting and Clinical Psychology, 73(5), 945e952.
Olatunji, B. O., & Wolitzky-Taylor, K. B. (2009). Anxiety sensitivity and the anxiety
disorders: a meta-analytic review and synthesis [Peer Reviewed]. Psychological
Bulletin, 135(6), 974e999. doi:10.1037/a0017428.
Orsillo, S. M., & Batten, S. V. (2005). Acceptance and commitment therapy in the
treatment of posttraumatic stress disorder. Behavior Modification, 29(1),
95e129.
Peterson, R. A., & Reiss, S. (1992). Anxiety sensitivity index manual (2nd ed.). Worthington, OH: International Diagnostic Systems.
Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. T. (2004). HLM 6:
Hierarchical linear and nonlinear modeling. Lincolnwood, IL: Scientific Software
International.
Raudenbush, S. W., & Liu, X.-F. (2001). Effects of study duration, frequency of
observation, and sample size on power in studies of group differences in
polynomial change. Psychological Methods, 6(4), 387e401.
Reiss, S., Peterson, R., Gursky, D., & McNally, R. (1986). Anxiety sensitivity, anxiety
frequency, and the prediction of fearfulness. Behavioural Research and Therapy,
24, 1e8.
Roemer, L., Orsillo, S. M., & Salters-Pedneault, K. (2008). Efficacy of an
acceptance-based behavior therapy for generalized anxiety disorder: evaluation
in a randomized controlled trial. Journal of Consulting and Clinical Psychology,
76(7), 1083e1089.
Segal, Z. V., Williams, J. M., & Teasdale, J. D. (2002). Mindfulness-based cognitive
therapy for depression: A new approach to preventing relapse. New York, NY:
Guilford Press.
Smits, J. A. J., Powers, M. B., Cho, Y., & Telch, M. J. (2004). Mechanism of change in
cognitive-behavioral treatment of panic disorder: evidence for the fear or fear
mediational hypothesis. Journal of Consulting and Clinical Psychology, 72(4),
646e652.
Smits, J. A. J., Rosenfeld, D., McDonald, R., & Telch, M. J. (2006). Cognitive
mechanisms of social anxiety reduction: an examination of specificity and
temporality. Journal of Consulting and Clinical Psychology, 74(6), 1203e1212.
Taylor, S., Koch, W. J., & McNally, R. (1992). How does anxiety sensitivity vary across
the anxiety disorders? Journal of Anxiety Disorders, 6, 249e259.
Tolin, D. F. (2010). Is cognitive behavioral therapy more effective than other
therapies?: a meta-analytic review. Clinical Psychology Review, 30(6), 710e720.
478
J.J. Arch et al. / Behaviour Research and Therapy 50 (2012) 469e478
Twohig, M. P., Hayes, S. C., & Masuda, A. (2006). Increasing willingness to
experience obsessions: acceptance and commitment therapy as a treatment for
obsessive-compulsive disorder. Behavior Therapy, 37(1), 3e13.
Twohig, M. P., Hayes, S. C., Plumb, J. C., Pruitt, L. D., Collins, A. B., Hazlett-Stevens, H.,
et al. (2010). A randomized clinical trial of acceptance and commitment therapy
vs. progressive relaxation training for obsessive compulsive disorder. Journal of
Consulting and Clinical Psychology, 78(5), 705e716.
Watson, D. (2005). Rethinking the mood and anxiety disorders: a quantitative
hierarchical model for DSM-V. Journal of Abnormal Psychology, 114(4), 522e536.
Watson, D., & Clark, L. A. (1991). The mood and anxiety symptom questionnaire.
Unpublished Manuscript. University of Iowa, Department of Psychology, Iowa City.
Watson, D., Clark, D. A., Weber, K., Assenheimer, J. S., Strauss, M. E., &
McCormick, R. A. (1995). Testing a tripartite model: II. Exploring the symptom
structure of anxiety and depression in student, adult, and patient samples.
Journal of Abnormal Psychology, 104, 15e25.
Watson, D., Weber, K., Assenheimer, J. S., Clark, L. A., Strauss, M. E., &
McCormick, R. A. (1995). Testing a tripartite model. I. Evaluating the convergent
and discriminant validity of anxiety and depression symptom. Journal of
Abnormal Psychology, 104, 3e14.
Zettle, R. D., & Hayes, S. C. (1986). Dysfunctional control by client verbal
behavior: the context of reason giving. The Analysis of Verbal Behavior, 4,
30e38.