Therapeutic Alliance As A Mediator of Change A Systematic Review and Evaluation of Research (Baier, Kline & Feeny, 2020)

Journal Pre-proof
Therapeutic alliance as a mediator of change: A systematic review

and evaluation of research
Allison L. Baier, Alexander C. Kline, Norah C. Feeny
PII: S0272-7358(20)30109-4
DOI: https://doi.org/10.1016/j.cpr.2020.101921
Reference: CPR 101921
To appear in: Clinical Psychology Review
Received date: 25 November 2019

Revised date: 6 August 2020
Accepted date: 14 September 2020
Please cite this article as: A.L. Baier, A.C. Kline and N.C. Feeny, Therapeutic alliance as a
mediator of change: A systematic review and evaluation of research, Clinical Psychology
Review (2019), https://doi.org/10.1016/j.cpr.2020.101921
This is a PDF file of an article that has undergone enhancements after acceptance, such
as the addition of a cover page and metadata, and formatting for readability, but it is
not yet the definitive version of record. This version will undergo additional copyediting,
typesetting and review before it is published in its final form, but we are providing this
version to give early visibility of the article. Please note that, during the production
process, errors may be discovered which could affect the content, and all legal disclaimers
that apply to the journal pertain.
© 2019 Published by Elsevier.

Journal Pre-proof
Running head: ALLIANCE AS A MEDIATOR OF CHANGE
Therapeutic Alliance as a Mediator of Change: A Systematic Review and Evaluation of Research
Allison L. Baiera, Alexander C. Klinea1, & Norah C. Feenya
a
PTSD Treatment and Research Program, Case Western Reserve University, Department of
Psychological Sciences, 11220 Bellflower Road, Cleveland, OH, USA, 44106-7123

1
Present address: VA San Diego Healthcare System, 3350 La Jolla Village Drive, MC116B, San
of
Diego, CA, USA 92161.
ro
-p
re
Preparation of this manuscript was supported by a grant to Dr. Feeny from the National Institute
of Mental Health (R01 MH066348). The funding source had no involvement in any aspect of this
lP
manuscript.
na
ur
Corresponding author: Allison L. Baier, Case Western Reserve University, Department of

Jo
Psychological Sciences, 11220 Bellflower Road, Cleveland, Ohio 44106. Email:
alb184@case.edu; Phone: 216-368-8934.
Abstract
The alliance-outcome relationship has been consistently linked to positive treatment outcomes
irrespective of psychotherapy modality. However, beyond its general links to favorable treatment
outcomes, it is less clear whether the alliance is a specific mediator of change and thus a possible
ALLIANCE AS A MEDIATOR OF Journal
CHANGE Pre-proof 2
mechanism underlying psychotherapy response. This systematic review evaluated research
examining the alliance as a potential mediator of symptom change, reviewing study
characteristics of 37 relevant articles examining the alliance-outcome relationship and the extent
to which these studies met recommended criteria for mechanistic research. Alliance mediated
therapeutic outcomes in 70.3% of the studies. We observed significant heterogeneity across
studies in terms of methodology, including timing of alliance assessment, study design,
constructs used in mediation models, and analytic approaches. Building on recent
of
methodological advancements, we propose directions for future research examining the putative
ro
mediational role of alliance, such as greater uniformity in and attention to study design and
-p
statistical methodology. This review highlights the importance of alliance in therapeutic change
and discusses how adhering to requirements for process research will improve our ability to more
re
precisely estimate how and to what extent alliance drives therapeutic change.
lP
Keywords: alliance, mediator, mechanism, psychotherapy, systematic review

na
ur
Jo
CHANGE Pre-proof 3
Therapeutic Alliance as a Mediator of Change: A Systematic Review and Evaluation of Research
The past several decades have seen the development and advancement of a range of
psychotherapies, including an increasing number of evidence-based psychotherapies with well-
defined treatment manuals and effectiveness in treating a range of mental health disorders
(Pincus & England, 2015). Despite this, dropout and nonresponse remain significant concerns
even among the most effective treatments, affecting an estimated 20-50% of patients (Nathan &
Gorman, 2015; Saxon, Firth, & Barkham, 2017; Wang et al., 2005). Improving patient adherence
of
and outcomes in psychotherapy will be aided by a greater understanding of how psychotherapy
ro
works. In line with this, the Institute of Medicine has called for research to identify elements of
-p
therapeutic change to better understand the mechanisms underlying treatment response
(Weissman, 2015). A more nuanced awareness of processes driving therapeutic change would
re
enable clinicians to further attend to the therapeutic elements actively contributing to treatment
lP
response or nonresponse, thereby optimizing treatments and ultimately providing more

na
personalized, effective care.
One hypothesized process underlying treatment response is the therapeutic alliance.

ur
Modern definitions of the term center on the alliance as a collaborative relationship between
Jo
therapist and patient that is influenced by the extent to which there is agreement on treatment
goals, a defined set of therapeutic tasks or processes to achieve the stated goals, and the
formation of a positive emotional bond (Bordin, 1979, 1994). Stronger alliance is consistently
associated with positive treatment outcomes across a range of psychotherapies as evidenced by
multiple meta-analyses on the subject, with fairly stable correlations between studies (Fluckiger,
Del Re, Wampold, & Horvath, 2018: r = .28, k = 295; Horvath & Bedi, 2002: r = .21, k = 100;
CHANGE Pre-proof 4
Horvath, Del Re, Fluckiger, & Symonds, 2011: r = .28, k = 190; Horvath & Symonds, 1991: r =
.28; k = 190; Martin, Garske, & Davis, 2000: r = .22, k = 79).
Although alliance has consistently been linked to better outcomes across psychotherapies,
there is ongoing debate with regard to the putative nature of alliance as an actual mechanism of
change. Some argue alliance is simply a precondition necessary for any successful therapy
(Hatcher & Barends, 2006; Raykos et al., 2014; Weck, Grikscheit, Jakob, Höfling, & Stangier,
2015). As a consequence, some methodologists argue alliance is a nonspecific treatment factor
of
that is important across all psychotherapies and thus is largely independent of psychotherapy
ro
“type”. In contrast, others argue that alliance is a specific treatment factor that drives therapeutic
-p
change itself and may be of greater significance in some psychotherapies (e.g., relational
therapies) over others (e.g., cognitive behavioral therapy; Siev, Huppert, & Chambless, 2009).
re
Whether a nonspecific or specific factor, therapeutic alliance is well studied, with over 306
lP
studies in the most recent meta-analysis on the alliance-outcome relationship alone (Fluckiger,
na
Del Re, Wampold, & Horvath, 2018). Although it is well established that strong alliance is
generally associated with better psychotherapy outcomes, the extent to which this process is
ur
itself specifically driving therapeutic change remains unclear. Randomized controlled trials
Jo
(RCTs) offer one way to examine the question of whether alliance is a specific or nonspecific
factor; if alliance is a specific factor, there should be treatment effects such that alliance plays a
more prominent role in one therapy over another whereas if alliance is a nonspecific factor,
studies would find no treatment effects. A resolution to this debate is important for better
understanding the therapeutic alliance’s impact on psychotherapy outcomes, as well as how this
process impacts outcomes. Additionally, a greater understanding of the alliance’s role in
psychotherapy will enable researchers to examine its potential interactive effects on other
CHANGE Pre-proof 5
possible mediators, thereby providing greater clarity regarding the mechanistic processes by
which treatments lead to change.
Requirements for a Mediator
Mediators help to explain why and how change is occurring and are considered the first
step in elucidating potential mechanisms of therapeutic change (Laurenceau, Hayes, & Feldman,
2007). Given that a subset of patients will drop out or not benefit from even the most effective
psychological treatments, identifying the processes underlying change will help implement
of
treatments in a manner emphasizing the most “essential” treatment components responsible for
ro
positive outcomes and also unveil ways to optimize such components for specific patients
-p
(Kraemer, Wilson, Fairburn, & Agras, 2002). Notably, while the terms ‘mediator’ and
‘mechanism’ are often used interchangeably, they are substantively different in that not all
re
mediators are mechanisms and researchers cannot necessarily make inferential conclusions from
lP
mediation analyses (Kazdin, 2007).

na
A mediator (variable M) is an intervening variable that statistically explains or accounts
for the relationship between two other variables: the independent variable (variable X) and
ur
dependent variable (variable Y) (MacKinnon, Fairchild, & Fritz, 2007). Three major approaches
Jo
are often used to assess mediation including: causal steps (Baron & Kenny, 1986; Judd & Kenny,
1981), difference in coefficients (Mackinnon & Dwyer, 1993), and product of coefficients
(Alwin & Hauser, 1975). The causal approach outlined by Baron and Kenny (1986) is most
widely used; however, it is limited by attenuated power and requires assumptions and study
requirements (e.g., normal distribution, large sample sizes) that can be hard to achieve in clinical
research (Hayes & Scharkow, 2013; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002;
Shrout & Bolger, 2002). Methodologists have put forth a number of alternative approaches to use
CHANGE Pre-proof 6
in conjunction with the major approaches or as new stand-alone methods. These include
bootstrap and other resampling methods that help deal with violations of the normality
assumption (Preacher & Hayes, 2004), the extension of single-mediator models to multiple
mediator models to examine the potential interactive effects of multiple variables (Hayes, 2013),
multilevel mediation models to handle hierarchical data (Bauer, Preacher, & Gil, 2006; Krull &
MacKinnon, 2001), and longitudinal mediation models to examine how variables change or
remain stable over the course of time (Cheong, MacKinnon, & Khoo, 2003; Curran & Bauer,
of
2011; Curran & Bollen, 2001; Curran, Lee, Howard, Lane, & MacCallum, 2012). While there
ro
are advantages and disadvantages to each method, multilevel longitudinal mediation methods
-p
offer the most robust statistical options for examining mediation in treatment data because these
models are equipped to handle the nested nature and time-course of the data. For a
re
comprehensive discussion of mediation analysis and the various pros and cons of different
lP
methods, readers are directed to MacKinnon and colleagues’ review (2007).

na
In addition to the mechanics of the mediation analysis, methodologists are increasingly
advocating for researchers to disentangle within-patient and between-patient effects in process

ur
research (Curran & Bauer, 2011). That is, how do the intraindividual and interindividual
Jo
variation in scores contribute to outcome over the course of time? Researchers often make
conclusions about within-patient processes from between-patient data. However, the between-
patient relationship between a process variable (e.g., alliance) and outcome (e.g., depression
symptoms) could in fact be a proxy for some other patient variable (e.g., diagnosis, personality),
creating difficulties when drawing inferential conclusions (Curran & Bauer, 2011). Thus, an
increased focus on longitudinal mediation methods that disaggregate the within-patient and
CHANGE Pre-proof 7
between-patient variation in alliance scores, is an important consideration for process research
that seeks to uncover why and how change occurs.
In sum, methodological approaches are integral to identifying and assessing proposed
mediators. Specifically, the precision and confidence of conclusions drawn are closely affected
by the types of analyses employed in studying this process.
Advances in Assessing Mediation
Mediators are an important first step in understanding hypothesized change processes;
of
however, the identification of a mediator does not necessarily explain the underlying cause of
ro
change (Kazdin & Nock, 2003; Laurenceau et al., 2007). In addition to the statistical requirement
-p
for mediation, Kazdin (2007) proposed six other criteria for drawing inferential conclusions from
studying change processes that can ultimately yield important clinical information about
re
psychotherapy. First, in addition to statistical mediation, researchers must be mindful of the
lP
temporal relationship (termed ‘timeline’; Kazdin, 2007) between the mediator and outcome, a
na
criterion that has increasingly been discussed and advocated (Johansson & Høglend, 2007;
Kazdin, 2007; Kraemer et al., 2002). That is, does the mediator precede and predict the outcome
ur
variable over the course of time? Second, researchers must rule out other explanations for the
Jo
observed relationship between the mediator and outcome. For example, what other competing
mediators need to be examined? Could the mediator in question simply be a proxy for some
other variable or patient trait (e.g., personality)? Or, if the mediator precedes outcome, might the
outcome variable also exert an influence over the hypothesized mediator (i.e., reverse causality)?
Kazdin (2007) refers to this criterion as ‘specificity’. Third, studies ought to manipulate the
hypothesized mediator through experimental designs (termed, ‘experimental manipulation’) to
demonstrate the relationships between the proposed mediator and outcome. Such experimental
CHANGE Pre-proof 8
methods would help determine Kazdin’s fourth criteria, ‘gradient’, wherein stronger “doses” of
the proposed mediator lead to greater change in outcome. Fifth, research findings must fit within
the broader scientific theory (‘plausibility/coherence’), and sixth, research must consistently
demonstrate a relationship between the mediator and outcome across replication studies,
including different patient populations and varying treatment conditions (‘consistency’; Kazdin,
2007). While all criteria are important to the study of mediators in the quest of identifying
mechanisms of change, some experts suggest that in statistical mediation, temporality,
of
specificity, and experimental manipulation should be considered the most important (Kazdin,
ro
2007; Kazdin & Nock, 2003). Accordingly, studies and reviews of mediation research have
-p
begun to focus on these particular criteria (e.g., Lemmens et al., 2017).
In addition to the criteria for studying mediators, researchers have put forth
re
recommended designs for process research, including randomized controlled trials (RCTs),
lP
comparisons with an adequate control group, sufficient power (e.g., sample size), and spaced
na
repeated measures (e.g., assessments at different timepoints throughout treatment) to allow for
adequate assessment of temporality (Kazdin, 2007; Kazdin & Nock, 2003; Laurenceau et al.,
ur
2007). Researchers have begun to evaluate the state of the literature with regard to these
Jo
recommendations for process research for both posttraumatic stress disorder (Cooper, Clifton, &
Feeny, 2017) and depression treatments (Lemmens et al., 2017). Clinically, such reviews provide
greater specificity regarding treatment processes that appear to be most effective in driving
therapeutic change. These reviews also shape future research on mediation by assessing
limitations in research to date and articulating research recommendations (e.g., study design,
statistical methodology) to better understand mediators in psychotherapy.

CHANGE Pre-proof 9
To date, meta-analyses have examined the alliance-outcome relationship as well as
potential moderators of the relationship, such as treatment type and patient characteristics
(Fluckiger et al., 2018; Horvath & Bedi, 2002; Horvath, Del Re, Flückiger, & Symonds, 2011;
Horvath & Symonds, 1991; Martin, Garske, & Davis, 2000). In a re-analysis (Flückiger, Del Re,
Wampold, Symonds, & Horvath, 2012) of data included in previous meta-analysis (Horvath et
al., 2011; k = 190 studies), the authors used multilevel, longitudinal meta-analytic procedures to
examine moderators of the alliance-outcome correlations over the course of therapy (e.g., design
of
characteristics, therapy type). While moving toward better understanding the nuance of the
ro
alliance-outcome relationship, like all other meta-analyses on the subject, the authors examined
-p
moderators of the alliance-outcome correlation. Beyond this relationship, the extent to which
alliance serves as a potential mediator of change in psychotherapy—and whether alliance may be

re
a specific or nonspecific factor—remains unclear, reflecting a critical next step in more precisely
lP
understanding the specific change processes occurring in treatment.

na
Aims of Current Review
The primary aim of this study was to provide a systematic review of research on the
ur
potential role of the therapeutic alliance as a mediator of change in psychotherapy. As therapy

Jo
change processes are better understood, clinicians will be better equipped to optimize these
“treatment drivers,” which ultimately should improve patient outcomes. Studies were selected
that examined the therapeutic alliance as a mediator between an independent variable and
treatment outcome with a statistical test of mediation (e.g., Baron & Kenny, 1986). In an effort to
better understand whether alliance is a specific factor on nonspecific factor, a broad range of in-
person, individual, outpatient psychotherapy was included. The resulting review presents the
CHANGE Pre-proof 10
characteristics of 37 studies examining alliance as a mediator of change and the extent to which
alliance is supported as a possible mediator of therapeutic change.
Method
Search Strategy
The search process occurred in two phases. First, PsycInfo was systematically searched
for potentially relevant papers published in peer-review journals. Limiters applied in the search
were publication date (January 1, 1980 and July 15, 2020), language (English only), and age
of
group (adulthood, defined as 18 years and older). The following search terms were used:
ro
“alliance” OR “therapeutic alliance” OR “helping alliance” OR “working alliance” AND
-p
“psychotherapy” OR “therapy” OR “cognitive (behavior(u)ral) therapy” OR “psychological
treatment(s)/intervention(s)” OR “Interpersonal (psycho)therapy” OR “psychodynamic/analytic

re
therapy” OR “client centered therapy” OR “acceptance commitment therapy” AND
lP
“mechanism” OR “mechanisms of change/action” OR “working mechanisms (of

na
psychotherapy)” OR “processes of therapy” OR “process research” OR “change” OR
“mediation” OR “mediator” OR “mediating effects”. Following the initial search, reference lists
ur
of prior meta-analyses of the alliance-outcome relationship (e.g., Horvath et al., 2011; Fluckiger
Jo
et al., 2018) were reviewed as well as reference lists of potentially suitable papers.
Inclusion Criteria
Studies were eligible for inclusion if they met the following criteria: (a) the study
consisted of adult patients; (b) the study was empirical and quantitative (i.e., reviews,
commentaries, theoretical essays, meta-analyses, systematic reviews, and qualitative studies
were excluded); (c) the study was not a case report; (d) patients received in-person, outpatient,
individual psychotherapy (i.e., group, inpatient, and telehealth modalities were excluded); (e) the
CHANGE Pre-proof 11
analytic method of the study included statistical mediation with the alliance as the hypothesized
mediating variable; (f) the study included a validated clinical symptom outcome measure as the
dependent variable in the mediation model(s) (e.g., Beck Depression Inventory; Beck, Steer,
Ball, & Ranieri, 1996); (g) the study used a validated measure of therapeutic alliance (e.g.,
Working Alliance Inventory; Hatcher & Gillaspy, 2006); and (h) the study was reported in
English. Inclusion criteria adhered to the precedent of prior alliance reviews examining in-
person, individual psychotherapy (e.g., Smith, Msetfi, & Golding, 2010) and reviews of process
of
mediation research (e.g., Lemmens et al., 2017). Given that the goal of this review was to
ro
evaluate the state of the literature, studies were not excluded for poor study quality (Cuijpers,
-p
van Straten, Bohlmeijer, Hollon, & Andersson, 2010; Hedges & Pigott, 2004); however, if more
than one study used an identical dataset, we chose to include the study with the strongest
re
methodology and study quality as defined by our coding criteria below.
lP
Coding Procedures
na
Key study characteristics and variables related to process research were coded by the first
author (blinded for review) and a trained, independent rater. Interrater reliability between coders
ur
was excellent ( = .93), and minimal differences in coding were resolved by consensus. Study
Jo
characteristics included: psychotherapy intervention type(s), primary diagnosis of the sample,
setting of study, validated measure of the therapeutic alliance, validated clinical symptom
outcome measure, statistical method(s) used to examine mediation, and main study findings.
Additionally, papers were assessed on whether or not they met key criteria for process research
following the methodology of Lemmens and colleagues (2017) including: the use of an RCT
design, use of a control group, sufficient sample size for mediation analyses (defined as n 40 in
line with other reviews evaluating process research), examination of multiple mediators within
CHANGE Pre-proof 12
one study, assessment of temporality (defined as 2 or more assessments of the therapeutic
alliance during the treatment phase that were examined over the course of time, not as an
average), and direct experimental manipulation of alliance. We chose to follow guidelines
outlined by Lemmens and colleagues in an effort to maintain consistency with the literature;
however, we also wanted to ensure maximum relevancy of criteria to the study of alliance as a
mediator of change. Consequently, we chose to collapse two criteria—RCT design and use of a
control group—into one criterion due to overlap (i.e., all studies in the review with a control
of
group were also an RCT). As previously stated, the direct comparison of two treatments can
ro
yield interesting information regarding whether alliance is a specific or nonspecific mediator of
-p
change. Additionally, we elected to add in a criterion regarding whether or not researchers
adequately disentangled within and between-patient effects specifically for the mediation
re
analysis, bringing the total criteria to six. Studies were rated as either meeting the criteria (1) or
lP
not (0).
na
Results
Study Selection
ur
Study selection followed PRISMA guidelines (Moher, Liberati, Tetzlaff, Altman, & The,
Jo
2009) and is presented in Figure 1. The initial PsycINFO search yielded 1,613 citations, which
were screened on the basis of title. Reference lists from meta-analyses and studies deemed
potentially eligible for inclusion were also reviewed, which provided an additional 59 citations
for review. Following this initial screening, a total of 479 studies with potential to meet inclusion
criteria were retained. Each abstract was then closely reviewed, followed by the full text (n =
192) if necessary to determine eligibility. A total of 442 studies were excluded, with the most
CHANGE Pre-proof 13
common reasons for exclusion including lack of empirical study and lack of mediation.
Ultimately, 37 studies met full inclusion criteria and are denoted with asterisks in the references.
Study Characteristics
Table 1 provides an overview of study characteristics and the extent to which each study
met predetermined criteria for process research. Thirty-seven studies were included in the review
with data from 5,530 patients. Studies were published between 1990 and 2020 with sample sizes
ranging between 20 to 646 (median = 103; M = 149; SD = 144.1). The majority of studies were
of
conducted in the United States (59.5% vs. 24.3% in Europe, and 16.2% in other parts of the
ro
world). Study settings were mixed with 10 studies (27.0%) utilizing data from training clinics or
-p
counseling centers and the remaining studies utilizing data from RCTs, hospitals, and/or
specialized treatment centers. Almost half (17 studies; 45.9%) were published in the last five
re
years, including the years 2015 to 2020. Eighteen of the 37 studies examined the alliance within
lP
the context of cognitive therapy (CT), cognitive behavioral therapy (CBT), or a combined
na
sample that included at least one arm of CT or CBT. Ten studies examined various forms of
psychotherapy, often in the context of community mental health clinics or university counseling
ur
centers. Other treatments examined either in isolation or, in a combined sample with CT/CBT
Jo
included Interpersonal Therapy (IPT, k = 4), Supportive Therapy (k = 2), Supportive Expressive
Therapy (SET, k = 3), Schema Focused Therapy (SFT, k = 2), Psychoanalytic Therapy (PA, k =
1), Psychodynamic Therapy (PD, k = 1), Cognitive Behavioral Analysis System of
Psychotherapy (CBASP, k = 1), Transference Focused Psychotherapy (TFP, k = 1), Functional
Analytic Psychotherapy (FAP, k = 1), Dialectical Behavior Therapy (DBT, k = 1), Unified
Protocol (UP, k = 1), Single Disorder Protocols (SDP, k = 1), and Return to Work Intervention
(RTW-I, k = 1). Six studies included a combined treatment or pharmacologic treatment.

CHANGE Pre-proof 14
The most common disorders studied were major depressive disorder (MDD; 15/37
studies, 40.1%) and mixed diagnostic samples (37.8%), followed by posttraumatic stress disorder
(PTSD; 5.4%), borderline personality disorder (BPD; 5.4%), chronic pain (2.7%), and bulimia
nervosa (2.7%), and exhaustion disorder (2.7%). Consequently, outcome measures varied
widely; the most commonly used outcome measure was the self-report Beck Depression
Inventory (Beck, Steer, Ball, & Ranieri, 1996), used in 11 of the 37 studies. The Working
Alliance Inventory or one of its short-form or revised versions (Hatcher & Gillaspy, 2006;
of
Horvath & Greenberg, 1989; Tracey & Kokotovic, 1989) was the alliance measure used most
ro
often (25/37 studies; 67.6%) followed by versions of the Helping Alliance (Alexander &
-p
Luborsky, 1986; Luborsky et al., 1996; Morgan, Luborsky, Crits-Christoph, Curtis, & Solomon,
1982), which was used in 4 studies (10.8%). In terms of who reported the quality of alliance, the
re
majority of studies (26/37) used patient ratings, six studies used independent observer ratings,
lP
and five used patient and therapist ratings (e.g., ran two separate models, one with patient ratings
na
and one with therapist ratings). The timing of alliance assessments also varied widely across
studies as can be seen in Table 1.

ur
The three earliest studies (Barber, Connolly, Crits-Christoph, Gladis, & Siqueland, 2009;
Jo
DeRubeis & Feeley, 1990; Feeley, DeRubeis, & Gelfand, 1999) used temporal correlations to
examine the relationship between alliance and residualized symptom change at different time
points in therapy. Only one of these three (Barber et al., 2000) found evidence that alliance
mediated symptom change. However, the methodology used in these first three studies limits the
ability to draw strong conclusions about the role of alliance on outcome. The most common
statistical method used to examine mediation was through a causal step approach utilizing linear
regressions (Baron & Kenny, 1986). Of the 13 studies using this method, eight relied on
CHANGE Pre-proof 15
bootstrapping methods to examine the size of the indirect effect (Preacher & Hayes, 2004) - a
method that can account for violation of normality assumptions (MacKinnon et al., 2007). Taken
together, the majority of studies using a causal step approach found support for alliance as a
mediator (10/13 studies).
Over the course of time, in line with advances in methodology, statistical methods used in
the identified studies became more sophisticated and included multilevel mediation (Krull &
MacKinnon, 2001), growth modeling (Cheong et al., 2003), and various forms of longitudinal
of
modeling (Curran & Bollen, 2001). Indeed, ten of the 17 studies published in the past five years
ro
(including 2015) used one of these methods (e.g., Falkenström, Ekeblad, & Holmqvist, 2016;
-p
Renner et al., 2018; Santoft et al., 2018). Of the 20 studies using more advanced statistical
methodology (e.g., multilevel and longitudinal modeling compared to simple mediation models),
re
the majority (15/20) found at least some evidence supporting alliance as a potential mediator. As
lP
presented in Table 1, studies were substantially different from one another in terms of design
na
(e.g., different populations, psychopathology, treatment-type) as well as the extent to which they
met requirements for process research. Furthermore, studies differed with how they handled
ur
missing data. Studies largely used available data (20/37), often in conjunction with imputation
Jo
methods and in line with analytic requirements or theoretical rationale (e.g., to be included in
analyses, patients needed at least “x” number of data points or patients needed to attend at least
“x” number of treatment sessions). Six studies used completer samples. The remaining studies
did not report how missing data were handled or, identifed an intent-to-treat approach (e.g., last
observation carried forward) and did not specify if they included patients with no data.
Conclusions regarding alliance as a mediator even among the most statistically advanced studies
should thus be interpreted in the context of the variability noted across studies.
CHANGE Pre-proof 16
Process Characteristics
For an overview on the number of studies meeting each of the requirements for process
research, readers are directed to Table 2. The majority of studies (81.1%) had sample sizes
greater than 40 patients and are thus likely sufficiently powered for appropriate inclusion in
systematic reviews or meta-analyses on the basis of sample size and power (Hedges & Pigott,
2004; Kazdin & Bass, 1989). Fewer than two-thirds of the studies (59.5%) included more than
two assessments of alliance, and only half examined other putative mediators in addition to the
of
alliance (51.4%). While 20 studies (54.5%) examined patients from RCTs, only 13 studies
ro
(48.1%) made use of the RCT design in examining differential treatment effects. Finally, seven
-p
studies (18.9%) adequately disentangled within and between-patient effects.
As noted, mediation analyses alone are not sufficient for drawing conclusions about
re
change processes. The strength of the argument for the mediator in question is proportional to the
lP
number of criteria met for process research (Kazdin, 2007). No study met all six criteria. Only
na
one study met five criteria. Falkenström and colleagues (2016) conducted an RCT comparing
CBT (n = 43) to IPT (n = 41) in a sample of patients with major depressive disorder. The authors
ur
utilized multilevel longitudinal models and additionally disaggregated within and between-
Jo
patient effects (Curran et al., 2012; Wang & Maxwell, 2015), finding a reciprocal relationship
between the alliance and change in depressive symptoms. That is, alliance predicted next session
change in depression scores and vice-versa, suggesting mutual influence of alliance and
symptoms. These results did not differ between treatments suggesting that the two-way
movement of alliance and symptom change was important across both treatment modalities
providing support for alliance as a common factor across treatments.

CHANGE Pre-proof 17
Six studies met four criteria and therefore seem to be promising with regard to meeting
process research requirements for understanding drivers of therapeutic change. Klein et al.
(2003) examined CBASP (n = 228) to CBASP plus antidepressant medication (n = 227) for the
treatment of depression. The authors utilized mixed effects growth modeling on treatment
initiators with baseline data to examine the temporal relationship between alliance and change in
depressive symptoms. The study found alliance at week 2 predicted subsequent improvement in
depressive symptoms after controlling for prior change. The authors found no evidence of
of
reverse causality (i.e., symptoms predicting improvements in alliance) and treatment condition
ro
did not moderate the findings.
-p
In a large study of primary care patients (N = 646), Falkenstrom and colleagues (2013)
used longitudinal multilevel models to examine whether alliance predicted symptom

re
improvement or vice versa. The authors additionally disentangled effects following procedures
lP
recommended by Curran and Bauer (2011) and found that within-patient alliance predicted next
na
session symptom change and vice versa. While this was a naturalistic sample, the authors
additionally examined whether the alliance-outcome relationship differed between treatment

ur
type, namely supportive, psychodynamic, or CBT finding no difference between groups lending
Jo
support for alliance as a nonspecific factor.
Sasso and colleagues (2016) re-examined data from the CT arm of a treatment trial for
depression by disaggregating within-patient and between-patient variance in alliance scores as
predictors of session to session symptom change early in treatment. The authors found that
neither within-patient nor between-patient variation in scores predicted subsequent symptom
change. The authors noted their limited sample size (N = 60) as a possible explanation for their
null findings.
CHANGE Pre-proof 18
In an RCT comparing SET (n = 49) to clinical management (e.g., supportive
interventions) and pharmacotherapy (n = 51) or clinical management and placebo (n = 49) for
patients diagnosed with depression (Zilcha-Mano et al., 2014), alliance temporally predicted
subsequent symptoms. The reverse relationship (symptom scores predicting next session alliance
scores) was not observed. No significant treatment interaction was observed suggesting alliance
was an important predictor of change across these treatment modalities.
Santoft et al. (2019) compared CBT (n = 40) to RTW-I (n = 42) for the treatment of
of
“exhaustion disorders” or burnout, a disorder found in the International Statistical Classification
ro
of Diseases and Related Health Problems (ICD-10; Organization, 2004). Over half of the sample
-p
(57.3%) met criteria for a comorbid disorder. The authors used a multilevel mediation model,
finding no association between the alliance and subsequent symptom changes over time.
re
Additionally, alliance did not mediate the relationship between treatment type and outcome.
lP
Finally, Gomez Penedo and colleagues (2020) compared the alliance as a predictor of
na
next session symptomatology in exposure-based cognitive therapy and CBT for patients with
depression. The authors used a hybrid random effect model finding both within-patient and
ur
between-patient alliance predicted next session symptomatology even when adjusting for
Jo
treatment condition.
Thus, of these seven studies scoring highest (4 or 5 out of 6 criteria) with regard to
meeting requirements for process research, findings are mixed. Five studies found evidence for
the mediating role of alliance whereas two studies did not. Additionally, four studies investigated
reciprocity (i.e., alliance predicting symptoms and symptoms predicting alliance); two found
support for this two-way movement between the alliance and symptom change. All five studies
finding support for alliance as a mediator of change also found support for alliance as a
CHANGE Pre-proof 19
nonspecific or “common” factor suggesting alliance is an important construct irrespective of
psychotherapy modality. Three studies (Falkenström et al., 2016; Sasso et al., 2016; Santoft et
al., 2019) had modest sample sizes under 100 patients which may have impacted findings;
however, all studies were adequately powered.
In addition, 7 studies (18.9%) met exactly half of the criteria for process research and 15
studies (40.5%) only met two criteria. Finally, 7 studies (18.9%) only met one of the six criteria.
While the combination of criteria met varied between studies, most met the sample size (30/37
of
studies) and temporality criteria (22/37 studies). It should be noted that the temporality criterion
ro
did not mandate testing for reciprocity or reverse causation, but only that the study included two
-p
or more assessments of alliance during the active phase of treatment. Taken together, 26 of the
37 studies (70.3%) found some evidence for the mediating role of alliance. Of note, only seven
re
studies (18.9%) could be considered of “highest quality” with respect to meeting at least 4
lP
criteria for process research. Given the small number of studies meeting these requirements,
na
results should be interpreted with caution.
Discussion
ur
Of all papers identified in this systematic review, the majority (70.3%; 26 of 37 studies)
Jo
found evidence for alliance as a mediator of change despite significant heterogeneity between
study designs, statistical analytic procedures, and overall quality. Although additional, more
targeted research is needed to more comprehensively unpack the alliance-outcome relationship,
results of this review reinforce that alliance likely plays an important role pantheoretically in
effective psychotherapy (Weck et al., 2015). The studies were critiqued with regard to the extent
to which they met six criteria for process research (Kazdin, 2007), in line with prior
methodology used to evaluate change mechanisms in clinical research (Lemmens et al., 2017)
CHANGE Pre-proof 20
while adapted to best meet this particular review. Only seven studies met four or more criteria
for process research pointing to clear future directions for the study of alliance as a change
mechanism. Nevertheless, an increasing number of research groups over the past five years made
use of robust analytic techniques that adequately deal with assumption violations, hierarchical
data, and longitudinal methods to include disaggregating within-patient and between-patient
effects, evidencing the promising advancements of recent research and methodological
approaches.
of
Given the positive alliance-outcome correlation that has consistently been observed in
ro
treatment research, alliance indeed seems to play some role in promoting symptom reduction,
-p
either mechanistically or by facilitating mechanistic processes. To answer the question of how
the alliance contributes to change, mediation analyses are needed coupled with robust research
re
designs that utilize RCTs to further investigate the question of whether the alliance is a specific
lP
or nonspecific factor of change. If the alliance is a nonspecific factor as the research to date
na
supports, then RCTs comparing different therapies should find no differences between treatments
in the role of alliance on treatment outcome. However, some continue to argue alliance as a
ur
specific change factor more important for certain psychotherapies (e.g., relational) than others
Jo
(e.g., exposure therapy). Additionally, studies should aim for large sample sizes to ensure
adequate power, multiple assessments of both validated alliance measures and validated outcome
measures to assess temporality, and the study of multiple constructs in concert. We further
acknowledge that given ethical and clinical demands of optimizing the therapeutic alliance in
treatment, the manipulation of alliance within a clinical research framework is challenging, if not
impossible. However, the ability to study if the magnitude of the mediator influences outcome is
a critical step in identifying mechanisms of change. It may be that the alliance is one purported
CHANGE Pre-proof 21
change process that will be difficult to unequivocally ascertain as a mechanism, whereas other
constructs (e.g., Socratic dialogue, exposure activities, behavioral activation) can be manipulated
ethically. However even when omitting this criterion, of the 37 included papers, less than half
(15 studies, 40.5%) met at least three of the other five requirements, evidencing the paucity of
literature adhering to stringent criteria for understanding mediators and drivers of treatment
change.
The 37 studies included in this review varied widely with regard to the extent to which
of
they met requirements for process research in addition to mediation method used and sample
ro
size. Ultimately, research aimed at uncovering possible change processes—including the
-p
alliance—must move toward inclusion of requirements for process research beginning with
mediation analyses. While the statistical methodology used in the included studies largely
re
improved over the course of time to account for more advanced models, future research will need
lP
to focus on using methodologically robust methods that are specific to the data in question. For
na
example, most mediation models examine linear changes; however, change is not always linear
(Hayes, Laurenceau, Feldman, Strauss, & Cardaciotto, 2007). Patients often show sudden gains
ur
(Jun, Zoellner, & Feeny, 2013) and experience ruptures and repairs in the alliance throughout the
Jo
course of treatment (McLaughlin, Keller, Feeny, Youngstrom, & Zoellner, 2014), which may not
be accurately reflected in linear models. Notably, we explicitly omitted such papers from this
review (e.g., Zilcha-Mano, Eubanks, Bloch-Elkouby, & Muran, 2020) for methodological
consistency and interpretation, as well as our focus in examining the links between outcomes and
alliance throughout the entire course of therapy. Thus, future studies should explore the use of
nonlinear and curvilinear models, which could help elucidate the temporal patterns associated
with change processes. Perhaps the alliance is critical early in the therapeutic process and less
CHANGE Pre-proof 22
susceptible to ruptures later on in treatment. In order to study temporal patterns, research must
make use of repeated measurement of alliance and symptoms over the course of treatment.
Identifying a time course of possible mediators, such as the alliance, will help clinicians better
understand when and where in treatment they might wish to direct their focus.
Identifying mediators ultimately relies on sound study designs including RCTs and
careful assessment of putative mediators to allow for an examination of temporality and
specificity. It will be important for future research to make use of these study designs. Better
of
understanding the mechanistic role of alliance will be assisted by research on alliance that is
ro
crafted during study design rather than being a secondary analysis of treatment data.
-p
Approaching the study of alliance from the outset will enable researchers to ensure inclusion of
important aspects of sound process research such as multiple assessments of alliance and other
re
putative change processes that might interact with alliance throughout the duration of treatment.
lP
Finally, as the alliance-outcome relationship is better understood, research should

na
continue to examine possible moderators of the relationship, such as therapist effects or patient
characteristics. Recent research, for example, has demonstrated that impact of alliance on
ur
outcomes may be more relevant for some patients than others. Specifically, the impact of alliance
Jo
on symptom change in patients receiving treatment for depression has been affected by
chronicity of depression, whereby the effects of alliance on outcome were greatest among
patients with fewer prior depressive episodes (e.g., Lorenzo-Luaces, DeRubeis, & Webb, 2014;
Lorenzo-Luaces et al., 2017). It is thus likely that alliance drives therapeutic change to a greater
degree for certain patients.
Clinically, cultivation of alliance should be prioritized at the earliest stages of treatment.
In addition to being broadly associated with optimal treatment outcomes, a stronger alliance
CHANGE Pre-proof 23
appears to also itself reflect an independent contributor to symptom reduction and likely be one
of the many processes driving change across therapy types, patient characteristics, and treatment
settings. Notably, while alliance appears to mediate change for some patients, it is also clear that
other treatment processes and techniques impact outcomes as well. Treatment process research
studying psychotherapy mechanisms has consistently implicated the role of specific processes on
symptom reduction (e.g., change in posttraumatic cognitions and fear reduction; Cooper et al.,
2017) in prolonged exposure for posttraumatic stress disorder). The current review suggests
of
however that, in addition to these specific treatment components, alliance itself often also
ro
contributes to therapeutic change for patients. It is thus possible alliance independently drives
-p
therapeutic change; however, it is likely more plausible that the alliance does not act in isolation
but rather facilitates other treatment processes (Lorenzo-Luaces & DeRubeis, 2018; Rothman,
re
2013). For example, the alliance between a patient and provider likely impacts the design and
lP
assignment of homework, creation of exposure exercises, or receptivity to Socratic dialogue.

na
Finally, in addition to alliance being particularly salient for symptom change among certain
patients, it is also likely that the alliance may be more relevant for the implementation of certain
ur
techniques rather than others (Tschacher, Junghan, & Pfammatter, 2014). Thus, perhaps the
Jo
quest to resolve the debate between whether specific or common factors are responsible for
therapeutic change is misguided. Research efforts that consider the complexity of the therapeutic
change process between specific and common factors and that capitalize on recent
methodological and statistical advances would further propel our scientific understanding of
change processes.
Given the importance of the alliance, clinicians should consider introducing routine and
systematic ways of monitoring the alliance such as with brief, validated patient-rated measures
CHANGE Pre-proof 24
(e.g., WAI; (Horvath & Greenberg, 1989). It is well-documented that there are benefits to
monitoring patient progress throughout the course of psychotherapy to track gains or setbacks
and make adjustments to the therapy as needed (Knaup, Koesters, Schoefer, Becker, & Puschner,
2009; Lambert & Lo Coco, 2013; Lambert & Shimokawa, 2011; Sapyta, Riemer, & Bickman,
2005). In addition to symptom monitoring, clinical outcomes across treatment are bolstered by
patient feedback related to the alliance (MacDonald, 2014; McClintock, Perlman, McCarrick,
Anderson, & Himawan, 2017; Norcross & Wampold, 2011). Systematically monitoring the
of
alliance would also assist clinicians’ attention to potential therapeutic ruptures, which have been
ro
shown not to negatively affect outcome so long as they are repaired (McLaughlin et al., 2014).
-p
Furthermore, studies have found that alliance scores are not inflated due to the presence of a
therapist or knowing that the scores would be reviewed by a therapist, which should relieve
re
clinician concerns regarding demand characteristics or social desirability of regular
lP
administration of alliance measures within psychotherapy (Reese et al., 2013).

na
Key strengths of this review include its systematic search adherent to PRISMA
guidelines, with a comprehensive examination of therapeutic alliance through mediation analyses

ur
and thus a mechanistic viewpoint. A diverse range of patients, primary diagnoses,

Jo
psychotherapies, and treatment settings were included enhancing the generalizability of the
findings. However, findings should be interpreted in the context of several limitations. As
presented in Table 1, study characteristics varied widely which should be considered in the
context of this review’s conclusions. Given the variability in methodological design implemented
across studies, we elected not to conduct a meta-analysis. We also note that the variables in the
relationships with alliance as a mediator (i.e., “X” and “Y” variables) were not uniform. The
heterogeneity of factors precluded the possibility of reliably and meaningfully evaluating the size
CHANGE Pre-proof 25
of effects related to alliance as a mediator. Advancements in methodology ultimately equate to
advancements in what can be reliably concluded regarding processes of change and, more
broadly, how treatments work (Zilcha-Mano, 2019). Notably, this review highlights significant
advancements in analytic approaches and methodology in recent years, with roughly half of
included studies published within the last five years. Recent papers have shifted from
correlational analyses taken at a snapshot during treatment to more advanced longitudinal models
that provide more precise estimates of the relationship between alliance and outcomes.
of
Additional progress in design and methodology will better illuminate the role that alliance plays
ro
in driving therapeutic change and enable future meta-analytic studies.
-p
Results of the current review suggests that alliance itself may be an independent driver of
therapeutic change. In the majority of studies included in the current review, alliance mediated
re
symptom reduction, supporting the alliance as a potential causal process, either independently or
lP
in conjunction with other change processes. This effect was observed across a broad range of
na
patients, disorders, and settings that were included in this review. Alliance likely impacts
psychotherapy in complex ways, reflecting the need for future targeted research to untangle these
ur
complex interactions and better elucidate when, how, for whom, and the extent to which alliance
Jo
serves as a mediator of outcome.

CHANGE Pre-proof 26
Acknowledgements
The authors wish to thank Alexandra Bowling for her diligent review of articles included in this
review.
of
ro
-p
re
lP
na
ur
Jo
CHANGE Pre-proof 27
References
Alexander, L. B., & Luborsky, L. (1986). The Penn Helping Alliance Scales. In L. S. Greenberg
& W. M. Pinsof (Eds.), The psychotherapeutic process: A research handbook. (pp. 325-
366). New York, NY: Guilford Press.
Alwin, D. F., & Hauser, R. M. (1975). The Decomposition of Effects in Path Analysis. American
Sociological Review, 40(1), 37-47. doi:10.2307/2094445
*Baldwin, S. A., Wampold, B. E., & Imel, Z. E. (2007). Untangling the alliance-outcome
of
correlation: Exploring the relative importance of therapist and patient variability in the
ro
alliance. Journal of Consulting and Clinical Psychology, 75(6), 842-852.
doi:10.1037/0022-006X.75.6.842
-p
*Barber, J. P., Connolly, M. B., Crits-Christoph, P., Gladis, L., & Siqueland, L. (2009). Alliance
re
predicts patients’ outcome beyond in-treatment change in symptoms. Personality
lP
Disorders: Theory, Research, and Treatment, S(1), 80-89. doi:10.1037/1949-2715.S.1.80

na
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social
psychological research: Conceptual, strategic, and statistical considerations. Journal of

ur
Personality and Social Psychology, 51(6), 1173-1182. doi:10.1037/0022-3514.51.6.1173

Jo
Bauer, D. J., Preacher, K. J., & Gil, K. M. (2006). Conceptualizing and testing random indirect
effects and moderated mediation in multilevel models: New procedures and
recommendations. Psychological Methods, 11(2), 142-163. doi:10.1037/1082-
989X.11.2.142
10.1037/1082-989X.11.2.142.supp (Supplemental)
CHANGE Pre-proof 28
Beck, A. T., Steer, R. A., Ball, R., & Ranieri, W. (1996). Comparison of Beck Depression
Inventories -IA and -II in psychiatric outpatients. J Pers Assess, 67(3), 588-597.
doi:10.1207/s15327752jpa6703_13
Bordin, E. S. (1979). The generalizability of the psychoanalytic concept of the working alliance.
Psychotherapy: Theory, Research & Practice, 16(3), 252-260. doi:10.1037/h0085885
Bordin, E. S. (1994). Theory and research on the therapeutic working alliance: New directions.
In A. O. Horvath & L. S. Greenberg (Eds.), The working alliance: Theory, research, and
of
practice. (pp. 13-37). Oxford, England: John Wiley & Sons.
ro
*Brattland, H., Koksvik, J. M., Burkeland, O., Klöckner, C. A., Lara-Cabrera, M. L., Miller, S.
-p
D., . . . Iversen, V. C. (2019). Does the working alliance mediate the effect of routine
re
outcome monitoring (ROM) and alliance feedback on psychotherapy outcomes? A
secondary analysis from a randomized clinical trial. Journal of Counseling Psychology,

lP
66(2), 234-246. doi:10.1037/cou0000320

na
*Burns, J. W., Nielson, W. R., Jensen, M. P., Heapy, A., Czlapinski, R., & Kerns, R. D. (2015).
Specific and general therapeutic mechanisms in cognitive behavioral treatment of chronic

ur
pain. Journal of Consulting and Clinical Psychology, 83(1), 1-11. doi:10.1037/a0037208

Jo
*Byrd, K. R., Patterson, C. L., & Turchik, J. A. (2010). Working alliance as a mediator of client
attachment dimensions and psychotherapy outcome. Psychotherapy: Theory, Research,
Practice, Training, 47(4), 631-636. doi:10.1037/a0022080
Cheong, J., MacKinnon, D. P., & Khoo, S. T. (2003). Investigation of mediational processes
using parallel process latent growth curve modeling. Structural Equation Modeling,
10(2), 238-262. doi:10.1207/S15328007SEM1002_5

CHANGE Pre-proof 29
Cooper, A. A., Clifton, E. G., & Feeny, N. C. (2017). An empirical review of potential mediators
and mechanisms of prolonged exposure therapy. Clinical Psychology Review, 56, 106-
121. doi:10.1016/j.cpr.2017.07.003
*Crits-Christoph, P., Gibbons, M. B. C., Hamilton, J., Ring-Kurtz, S., & Gallop, R. (2011). The
dependability of alliance assessments: The alliance–outcome correlation is larger than
you might think. Journal of Consulting and Clinical Psychology, 79(3), 267-278.
doi:10.1037/a0023668
of
Cuijpers, P., van Straten, A., Bohlmeijer, E., Hollon, S. D., & Andersson, G. (2010). The effects
ro
of psychotherapy for adult depression are overestimated: A meta-analysis of study quality
-p
and effect size. Psychological Medicine, 40(2), 211-223.
doi:10.1017/S0033291709006114
re
Curran, P. J., & Bauer, D. J. (2011). The disaggregation of within-person and between-person
lP
effects in longitudinal models of change. Annual Review of Psychology, 62, 583-619.

na
doi:10.1146/annurev.psych.093008.100356
Curran, P. J., & Bollen, K. A. (2001). The best of both worlds: Combining autoregressive and
ur
latent curve models. In L. M. Collins & A. G. Sayer (Eds.), New methods for the analysis
Jo
of change. (pp. 107-135). Washington, DC: American Psychological Association.
Curran, P. J., Lee, T., Howard, A. L., Lane, S., & MacCallum, R. (2012). Disaggregating within-
person and between-person effects in multilevel and structural equation growth models.
In J. R. Harring, G. R. Hancock, J. R. Harring, & G. R. Hancock (Eds.), Advances in
longitudinal methods in the social and behavioral sciences. (pp. 217-253). Charlotte, NC,
US: IAP Information Age Publishing.

CHANGE Pre-proof 30
*DeRubeis, R. J., & Feeley, M. (1990). Determinants of change in cognitive therapy for
depression. Cognitive Therapy and Research, 14(5), 469-482. doi:10.1007/BF01172968
*Falkenström, F., Ekeblad, A., & Holmqvist, R. (2016). Improvement of the working alliance in
one treatment session predicts improvement of depressive symptoms by the next session.
Journal of Consulting and Clinical Psychology, 84(8), 738-751. doi:10.1037/ccp0000119
*Falkenström, F., Granström, F., & Holmqvist, R. (2013). Therapeutic alliance predicts
symptomatic improvement session by session. Journal of Counseling Psychology, 60(3),
of
317-328. doi:10.1037/a0032258
ro
*Feeley, M., DeRubeis, R. J., & Gelfand, L. A. (1999). The temporal relation of adherence and
-p
alliance to symptom change in cognitive therapy for depression. J Consult Clin Psychol,
67(4), 578-582.
re
Fluckiger, C., Del Re, A. C., Wampold, B. E., & Horvath, A. O. (2018). The alliance in adult
lP
psychotherapy: A meta-analytic synthesis. Psychotherapy (Chic), 55(4), 316-340.

na
doi:10.1037/pst0000172
Flückiger, C., Del Re, A. C., Wampold, B. E., Symonds, D., & Horvath, A. O. (2012). How
ur
Central Is the Alliance in Psychotherapy? A Multilevel. Journal of Counseling

Jo
Psychology, 59(1), 10-17. doi:10.1037/A0025749
*Forbes, D., Parslow, R., Creamer, M., Allen, N., McHugh, T., & Hopwood, M. (2008).
Mechanisms of anger and treatment outcome in combat veterans with posttraumatic
stress disorder. Journal of Traumatic Stress, 21(2), 142-149. doi:10.1002/jts.20315
*Gómez Penedo, J. M., Babl, A., Krieger, T., Heinonen, E., Flückiger, C., & grosse Holtforth,
M. (2020). Interpersonal agency as predictor of the within-patient alliance effects on

CHANGE Pre-proof 31
depression severity. Journal of Consulting and Clinical Psychology, 88(4), 338-349.
doi:10.1037/ccp0000475
Hatcher, R. L., & Barends, A. W. (2006). How a return to theory could help alliance research.
Psychotherapy: Theory, Research, Practice, Training, 43(3), 292-299. doi:10.1037/0033-
3204.43.3.292
Hatcher, R. L., & Gillaspy, J. A. (2006). Development and validation of a revised short version
of the Working Alliance Inventory. Psychotherapy Research, 16(1), 12-25.
of
doi:10.1080/10503300500352500
ro
Hayes, & Scharkow, M. (2013). The relative trustworthiness of inferential tests of the indirect
-p
effect in statistical mediation analysis: Does method really matter? Psychological
Science, 24(10), 1918-1927. doi:10.1177/0956797613480187

re
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A
lP
regression-based approach. New York, NY: Guilford Press.

na
Hayes, A. M., Laurenceau, J.-P., Feldman, G., Strauss, J. L., & Cardaciotto, L. (2007). Change is
not always linear: The study of nonlinear and discontinuous patterns of change in
ur
psychotherapy. Clinical Psychology Review, 27(6), 715-723.

Jo
doi:10.1016/j.cpr.2007.01.008
Hedges, L. V., & Pigott, T. D. (2004). The Power of Statistical Tests for Moderators in Meta-
Analysis. Psychological Methods, 9(4), 426-445. doi:10.1037/1082-989X.9.4.426
*Hirsh, J. B., Quilty, L. C., Bagby, R. M., & McMain, S. F. (2012). The relationship between
agreeableness and the development of the working alliance in patients with borderline
personality disorder. Journal of Personality Disorders, 26(4), 616-627.
doi:10.1521/pedi.2012.26.4.616
CHANGE Pre-proof 32
Horvath, A. O., & Bedi, R. P. (2002). The alliance. In J. C. Norcross (Ed.), Psychotherapy
relationships that work: Evidence-based responsiveness (pp. 37-69). New York, NY:
Oxford University Press.
Horvath, A. O., Del Re, A. C., Flückiger, C., & Symonds, D. (2011). Alliance in individual
psychotherapy. Psychotherapy, 48(1), 9-16. doi:10.1037/a0022186
Horvath, A. O., & Greenberg, L. S. (1989). Development and validation of the Working Alliance
Inventory. Journal of Counseling Psychology, 36(2), 223-233. doi:10.1037/0022-
of
0167.36.2.223
ro
Horvath, A. O., & Symonds, B. D. (1991). Relation between working alliance and outcome in
-p
psychotherapy: A meta-analysis. Journal of Counseling Psychology, 38(2), 139-149.
doi:10.1037/0022-0167.38.2.139
re
Johansson, P., & Høglend, P. (2007). Identifying mechanisms of change in psychotherapy:
lP
Mediators of treatment outcome. Clinical Psychology & Psychotherapy, 14(1), 1-9.

na
doi:10.1002/cpp.514
Judd, C. M., & Kenny, D. A. (1981). Process Analysis: Estimating Mediation in Treatment
ur
Evaluations. Evaluation Review, 5(5), 602-619. doi:10.1177/0193841X8100500502

Jo
Jun, J. J., Zoellner, L. A., & Feeny, N. C. (2013). Sudden gains in prolonged exposure and
sertraline for chronic PTSD. Depression and Anxiety, 30(7), 607-613.
doi:10.1002/da.22119
Kazdin, A. E. (2007). Mediators and mechanisms of change in psychotherapy research. Annual
Review of Clinical Psychology, 3, 1-27. doi:10.1146/annurev.clinpsy.3.022806.091432

CHANGE Pre-proof 33
Kazdin, A. E., & Bass, D. (1989). Power to detect differences between alternative treatments in
comparative psychotherapy outcome research. Journal of Consulting and Clinical
Psychology, 57(1), 138-147. doi:10.1037/0022-006X.57.1.138
Kazdin, A. E., & Nock, M. K. (2003). Delineating mechanisms of change in child and adolescent
therapy: methodological issues and research recommendations. Journal of Child
Psychology and Psychiatry, 44(8), 1116-1129. doi:10.1111/1469-7610.00195
*Klein, D. N., Schwartz, J. E., Santiago, N. J., Vivian, D., Vocisano, C., Castonguay, L. G., . . .
of
Keller, M. B. (2003). Therapeutic Alliance in Depression Treatment: Controlling for
ro
Prior Change and Patient Characteristics. Journal of Consulting and Clinical Psychology,
-p
71(6), 997-1006. doi:10.1037/0022-006X.71.6.997
*Klug, G., Zimmermann, J., & Huber, D. (2016). Outcome trajectories and mediation in
re
psychotherapeutic treatments of major depression. Journal of the American
lP
Psychoanalytic Association, 64(2), 307-343. doi:10.1177/0003065116644742

na
Knaup, C., Koesters, M., Schoefer, D., Becker, T., & Puschner, B. (2009). Effect of feedback of
treatment outcome in specialist mental healthcare: Meta-analysis. The British Journal of

ur
Psychiatry, 195(1), 15-22. doi:10.1192/bjp.bp.108.053967

Jo
Kraemer, H. C., Wilson, G. T., Fairburn, C. G., & Agras, W. S. (2002). Mediators and
moderators of treatment effects in randomized clinical trials. Archives of General
Psychiatry, 59(10), 877-883. doi:10.1001/archpsyc.59.10.877
Krull, J. L., & MacKinnon, D. P. (2001). Multilevel modeling of individual and group level
mediated effects. Multivariate Behavioral Research, 36(2), 249-277.
doi:10.1207/S15327906MBR3602_06
CHANGE Pre-proof 34
*Kushner, S. C., Quilty, L. C., Uliaszek, A. A., McBride, C., & Bagby, R. M. (2016).
Therapeutic alliance mediates the association between personality and treatment outcome
in patients with major depressive disorder. Journal of Affective Disorders, 201, 137-144.
doi:10.1016/j.jad.2016.05.016
Lambert, M. J., & Lo Coco, G. (2013). Simple methods for enhancing patient outcome in routine
care: Measuring, monitoring, and feedback. Research in Psychotherapy:
Psychopathology, Process and Outcome, 16(2), 93-101. doi:10.4081/ripppo.2013.147
of
Lambert, M. J., & Shimokawa, K. (2011). Collecting client feedback. Psychotherapy, 48(1), 72-
ro
79. doi:10.1037/a0022238
-p
Laurenceau, J.-P., Hayes, A. M., & Feldman, G. C. (2007). Some methodological and statistical
issues in the study of change processes in psychotherapy. Clinical Psychology Review,

re
27(6), 682-695. doi:10.1016/j.cpr.2007.01.007
lP
*Lawson, D. M., Stulmaker, H., & Tinsley, K. (2017). Therapeutic alliance, interpersonal
na
relations, and trauma symptoms: Examining a mediation model of women with childhood
abuse histories. Journal of Aggression, Maltreatment & Trauma, 26(8), 861-878.

ur
doi:10.1080/10926771.2017.1331941
Jo
*Leibovich, L., Front, O., McCarthy, K. S., & Zilcha-Mano, S. (2020). How do supportive
techniques bring about therapeutic change: The role of therapeutic alliance as a potential
mediator. Psychotherapy, 57(2), 151-159. doi:10.1037/pst0000253
Lemmens, L. H. J. M., Galindo-Garre, F., Arntz, A., Peeters, F., Hollon, S. D., DeRubeis, R. J.,
& Huibers, M. J. H. (2017). Exploring mechanisms of change in cognitive therapy and
interpersonal psychotherapy for adult depression. Behaviour Research and Therapy, 94,
81-92. doi:10.1016/j.brat.2017.05.005
CHANGE Pre-proof 35
Lorenzo-Luaces, L., & DeRubeis, R. J. (2018). Miles to Go Before We Sleep: Advancing the
Understanding of Psychotherapy by Modeling Complex Processes. Cognitive Therapy
and Research, 42(2), 212-217. doi:10.1007/s10608-018-9893-x
Lorenzo-Luaces, L., DeRubeis, R. J., & Webb, C. A. (2014). Client characteristics as moderators
of the relation between the therapeutic alliance and outcome in cognitive therapy for
depression. J Consult Clin Psychol, 82(2), 368-373. doi:10.1037/a0035994
Lorenzo-Luaces, L., Driessen, E., DeRubeis, R. J., Van, H. L., Keefe, J. R., Hendriksen, M., &
of
Dekker, J. (2017). Moderation of the Alliance-Outcome Association by Prior Depressive
ro
Episodes: Differential Effects in Cognitive-Behavioral Therapy and Short-Term
-p
Psychodynamic Supportive Psychotherapy. Behav Ther, 48(5), 581-595.
doi:10.1016/j.beth.2016.11.011
re
Luborsky, L., Barber, J. P., Siqueland, L., Johnson, S., Najavits, L. M., Frank, A., & Daley, D.
lP
(1996). The revised Helping Alliance questionnaire (HAq-II): Psychometric properties.

na
Journal of Psychotherapy Practice & Research, 5(3), 260-271.
MacDonald, J. (2014). 'Formal' feedback in psychotherapy as psychoanalytic technique.

ur
Psychodynamic Practice: Individuals, Groups and Organisations, 20(2), 154-163.

Jo
doi:10.1080/14753634.2013.771566
Mackinnon, D. P., & Dwyer, J. H. (1993). Estimating Mediated Effects in Prevention Studies.
Evaluation Review, 17(2), 144-158. doi:10.1177/0193841X9301700202
MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual Review of
Psychology, 58, 593-614. doi:10.1146/annurev.psych.58.110405.085542

CHANGE Pre-proof 36
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A
comparison of methods to test mediation and other intervening variable effects.
Psychological Methods, 7(1), 83-104. doi:10.1037/1082-989X.7.1.83
*Maitland, D. W. M., Petts, R. A., Knott, L. E., Briggs, C. A., Moore, J. A., & Gaynor, S. T.
(2016). A randomized controlled trial of functional analytic psychotherapy versus
watchful waiting: Enhancing social connectedness and reducing anxiety and avoidance.
Behavior Analysis: Research and Practice, 16(3), 103-122. doi:10.1037/bar0000051
of
Martin, D. J., Garske, J. P., & Davis, M. K. (2000). Relation of the therapeutic alliance with
ro
outcome and other variables: A meta-analytic review. Journal of Consulting and Clinical
-p
Psychology, 68(3), 438-450. doi:10.1037/0022-006X.68.3.438
*McClintock, A. S., Anderson, T., & Petrarca, A. (2015). Treatment expectations, alliance,
re
session positivity, and outcome: An investigation of a three‐path mediation model.
lP
Journal of Clinical Psychology, 71(1), 41-49. doi:10.1002/jclp.22119

na
McClintock, A. S., Perlman, M. R., McCarrick, S. M., Anderson, T., & Himawan, L. (2017).
Enhancing psychotherapy process with common factors feedback: A randomized, clinical

ur
trial. Journal of Counseling Psychology, 64(3), 247-260. doi:10.1037/cou0000188

Jo
McLaughlin, A. A., Keller, S. M., Feeny, N. C., Youngstrom, E. A., & Zoellner, L. A. (2014).
Patterns of therapeutic alliance: Rupture–repair episodes in prolonged exposure for
posttraumatic stress disorder. Journal of Consulting and Clinical Psychology, 82(1), 112-
121. doi:10.1037/a0034696
Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G., & The, P. G. (2009). Preferred Reporting
Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLOS
Medicine, 6(7), e1000097. doi:10.1371/journal.pmed.1000097

CHANGE Pre-proof 37
Morgan, R., Luborsky, L., Crits-Christoph, P., Curtis, H., & Solomon, J. (1982). Predicting the
outcomes of psychotherapy by the Penn Helping Alliance Rating Method. Archives of
General Psychiatry, 39(4), 397-402. doi:10.1001/archpsyc.1982.04290040013002
Nathan, P. E., & Gorman, J. M. (2015). A guide to treatments that work., 4th ed. New York, NY,
US: Oxford University Press.
Norcross, J. C., & Wampold, B. E. (2011). Evidence-based therapy relationships: Research
conclusions and clinical practices. Psychotherapy, 48(1), 98-102. doi:10.1037/a0022161
of
Organization, W. H. (2004). ICD-10: International statistical classification of diseases and
ro
related health problems (tenth revision, 2nd ed.): World Health Organization.
-p
*Owen, J., Imel, Z., Tao, K. W., Wampold, B., Smith, A., & Rodolfa, E. (2011). Cultural
ruptures in short-term therapy: Working alliance as a mediator between clients'

re
perceptions of microaggressions and therapy outcomes. Counselling & Psychotherapy
lP
Research, 11(3), 204-212. doi:10.1080/14733145.2010.491551

na
*Patterson, C. L., Anderson, T., & Wei, C. (2014). Clients' pretreatment role expectations, the
therapeutic alliance, and clinical outcomes in outpatient therapy. J Clin Psychol, 70(7),
ur
673-680. doi:10.1002/jclp.22054
Jo
Pincus, H. A., & England, M. J. (2015). Improving the Quality of Psychosocial Interventions for
Mental and Substance Use Disorders: A Report From the IOMPsychosocial Interventions
for Mental Health and Substance Use DisordersPsychosocial Interventions for Mental
Health and Substance Use Disorders. Jama, 314(12), 1227-1228.
doi:10.1001/jama.2015.9212
CHANGE Pre-proof 38
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects
in simple mediation models. Behavior Research Methods, Instruments & Computers,
36(4), 717-731. doi:10.3758/BF03206553
Raykos, B. C., McEvoy, P. M., Erceg-Hurn, D., Byrne, S. M., Fursland, A., & Nathan, P. (2014).
Therapeutic alliance in enhanced cognitive behavioural therapy for bulimia nervosa:
Probably necessary but definitely insufficient. Behaviour Research and Therapy, 57, 65-
71. doi:10.1016/j.brat.2014.04.004
of
Reese, R. J., Gillaspy, J. A., Jr., Owen, J. J., Flora, K. L., Cunningham, L. C., Archie, D., &
ro
Marsden, T. (2013). The influence of demand characteristics and social desirability on
-p
clients’ ratings of the therapeutic alliance. Journal of Clinical Psychology, 69(7), 696-
709. doi:10.1002/jclp.21946
re
*Renner, F., DeRubeis, R., Arntz, A., Peeters, F., Lobbestael, J., & Huibers, M. J. H. (2018).
lP
Exploring mechanisms of change in schema therapy for chronic depression. Journal of

na
Behavior Therapy and Experimental Psychiatry, 58, 97-105.
doi:10.1016/j.jbtep.2017.10.002
ur
Rothman, A. J. (2013). Exploring connections between moderators and mediators: commentary

Jo
on subgroup analyses in intervention research. Prev Sci, 14(2), 189-192.
doi:10.1007/s11121-012-0333-y
*Rubel, J. A., Hilpert, P., Wolfer, C., Held, J., Vîslă, A., & Flückiger, C. (2019). The working
alliance in manualized CBT for generalized anxiety disorder: Does it lead to change and
does the effect vary depending on manual implementation flexibility? Journal of
Consulting and Clinical Psychology. doi:10.1037/ccp0000433
10.1037/ccp0000433.supp (Supplemental)
CHANGE Pre-proof 39
*Santoft, F., Salomonsson, S., Hesser, H., Lindsäter, E., Ljótsson, B., Lekander, M., . . .
Hedman-Lagerlöf, E. (2019). Mediators of change in cognitive behavior therapy for
clinical burnout. Behavior Therapy, 50(3), 475-488. doi:10.1016/j.beth.2018.08.005
Sapyta, J., Riemer, M., & Bickman, L. (2005). Feedback to Clinicians: Theory, Research, and
Practice. Journal of Clinical Psychology, 61(2), 145-153. doi:10.1002/jclp.20107
*Sasso, K. E., Strunk, D. R., Braun, J. D., DeRubeis, R. J., & Brotman, M. A. (2016). A re-
examination of process–outcome relations in cognitive therapy for depression:
of
Disaggregating within-patient and between-patient effects. Psychotherapy Research,
ro
26(4), 387-398. doi:10.1080/10503307.2015.1026423
-p
*Sauer-Zavala, S., Boswell, J. F., Bentley, K. H., Thompson-Hollands, J., Farchione, T. J., &
Barlow, D. H. (2018). Expectancies, working alliance, and outcome in transdiagnostic

re
and single diagnosis treatment for anxiety disorders: An investigation of mediation.
lP
Cognitive Therapy and Research, 42(2), 135-145. doi:10.1007/s10608-017-9855-8

na
Saxon, D., Firth, N., & Barkham, M. (2017). The relationship between therapist effects and
therapy delivery factors: Therapy modality, dosage, and non-completion. Administration

ur
and Policy in Mental Health and Mental Health Services Research, 44(5), 705-715.
Jo
doi:10.1007/s10488-016-0750-5
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New
procedures and recommendations. Psychological Methods, 7(4), 422-445.
doi:10.1037/1082-989X.7.4.422
Siev, J., Huppert, J. D., & Chambless, D. L. (2009). The Dodo Bird, treatment technique, and
disseminating empirically supported treatments. the Behavior Therapist, 32(4), 71-76.

CHANGE Pre-proof 40
Smith, A. E. M., Msetfi, R. M., & Golding, L. (2010). Client self rated adult attachment patterns
and the therapeutic alliance: A systematic review. Clinical Psychology Review, 30(3),
326-337. doi:10.1016/j.cpr.2009.12.007
*Spinhoven, P., Giesen-Bloo, J., van Dyck, R., Kooiman, K., & Arntz, A. (2007). The
therapeutic alliance in schema-focused therapy and transference-focused psychotherapy
for borderline personality disorder. Journal of Consulting and Clinical Psychology,
75(1), 104-115. doi:10.1037/0022-006X.75.1.104
of
*Strunk, D. R., Cooper, A. A., Ryan, E. T., DeRubeis, R. J., & Hollon, S. D. (2012). The process
ro
of change in cognitive therapy for depression when combined with antidepressant
-p
medication: Predictors of early intersession symptom gains. Journal of Consulting and
Clinical Psychology, 80(5), 730-738. doi:10.1037/a0029281

re
10.1037/a0029281.supp (Supplemental)
lP
*Sullivan, J. M., Lawson, D. M., & Akay-Sullivan, S. (2020). Insecure attachment and
na
therapeutic bond as mediators of social, relational, and social distress and interpersonal
problems in adult females with childhood sexual abuse history. Journal of Child Sexual
ur
Abuse: Research, Treatment, & Program Innovations for Victims, Survivors, &
Jo
Offenders.
Tracey, T. J., & Kokotovic, A. M. (1989). Factor structure of the Working Alliance Inventory.
Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1(3), 207-
210. doi:10.1037/1040-3590.1.3.207
Tschacher, W., Junghan, U. M., & Pfammatter, M. (2014). Towards a taxonomy of common
factors in psychotherapy—Results of an expert survey. Clinical Psychology &
Psychotherapy, 21(1), 82-96. doi:10.1002/cpp.1822

CHANGE Pre-proof 41
Wang, L., & Maxwell, S. E. (2015). On disaggregating between-person and within-person
effects with longitudinal data using multilevel models. Psychological Methods, 20(1), 63-
83. doi:10.1037/met0000030
Wang, P. S., Lane, M., Olfson, M., Pincus, H. A., Wells, K. B., & Kessler, R. C. (2005). Twelve-
month use of mental health services in the United States: results from the National
Comorbidity Survey Replication. Arch Gen Psychiatry, 62(6), 629-640.
doi:10.1001/archpsyc.62.6.629
of
*Webb, C. A., DeRubeis, R. J., Amsterdam, J. D., Shelton, R. C., Hollon, S. D., & Dimidjian, S.
ro
(2011). Two aspects of the therapeutic alliance: differential relations with depressive
-p
symptom change. J Consult Clin Psychol, 79(3), 279-283. doi:10.1037/a0023252
Weck, F., Grikscheit, F., Jakob, M., Höfling, V., & Stangier, U. (2015). Treatment failure in
re
cognitive‐behavioural therapy: Therapeutic alliance as a precondition for an adherent and
lP
competent implementation of techniques. British Journal of Clinical Psychology, 54(1),

na
91-108. doi:10.1111/bjc.12063
Weissman, M. M. (2015). The institute of medicine (IOM) sets a framework for evidence-baed
ur
standards for psychotherapy. Depression and Anxiety, 32(11), 787-789.

Jo
doi:10.1002/da.22434
*Wilson, G. T., Fairburn, C. C., Agras, W. S., Walsh, B. T., & Kraemer, H. (2002). Cognitive-
behavioral therapy for bulimia nervosa: Time course and mechanisms of change. Journal
of Consulting and Clinical Psychology, 70(2), 267-274. doi:10.1037/0022-006X.70.2.267
*Xu, H., & Tracey, T. J. G. (2015). Reciprocal influence model of working alliance and
therapeutic outcome over individual therapy course. Journal of Counseling Psychology,
62(3), 351-359. doi:10.1037/cou0000089

CHANGE Pre-proof 42
*Yoo, S.-K., Hong, S., Sohn, N., & O'Brien, K. M. (2014). Working alliance as a mediator and
moderator between expectations for counseling success and counseling outcome among
Korean clients. Asia Pacific Education Review, 15(2), 271-281. doi:10.1007/s12564-014-
9320-2
*Zilcha-Mano, S., & Errázuriz, P. (2017). Early development of mechanisms of change as a
predictor of subsequent change and treatment outcome: The case of working alliance.
Journal of Consulting and Clinical Psychology, 85(5), 508-520. doi:10.1037/ccp0000192
of
10.1037/ccp0000192.supp (Supplemental)
ro
*Zilcha-Mano, S., Dinger, U., McCarthy, K. S., & Barber, J. P. (2014). Does alliance predict
-p
symptoms throughout treatment, or is it the other way around? J Consult Clin Psychol,
82(6), 931-935. doi:10.1037/a0035141

re
Zilcha-Mano, S. (2019). Major developments in methods addressing for whom psychotherapy
lP
may work and why. Psychother Res, 29(6), 693-708.

na
doi:10.1080/10503307.2018.1429691
Zilcha-Mano, S., Eubanks, C. F., Bloch-Elkouby, S., & Muran, J. C. (2020). Can we agree we
ur
just had a rupture? Patient-therapist congruence on ruptures and its effects on outcome in
Jo
brief relational therapy versus cognitive-behavioral therapy. Journal of Counseling
Psychology, 67(3), 315-325. doi:10.1037/cou0000400
10.1037/cou0000400.supp (Supplemental)
*Zuroff, D. C., Blatt, S. J., Sotsky, S. M., Krupnick, J. L., Martin, D. J., Sanislow, C. A., 3rd, &
Simmens, S. (2000). Relation of therapeutic alliance and perfectionism to outcome in
brief outpatient treatment of depression. J Consult Clin Psychol, 68(1), 114-124.

ALLIANCE AS A MEDIATOR OF CHANGE Journal Pre-proof 43
Table 1
Characteristics and results of 37 identified studies aimed at examining the potential role of the therapeutic alliance as a mediator of treatment outcomes and the
extent to which they meet requirements for process research.
Study Characteristics and Results Requirements for Process Research

Primary Setting & Diagnosis Alliance Timepoint of Statistical Main Finding(s) RCT Disentan N 40 Multiple Temporal Manipula
Author, Intervention(s)* Measure Alliance Measure Mediation gled Mediator ity tion
Date
DeRubeis
et al.,
Outpatient Center;
CT (n = 25)
MDD PHAS
(observer)
Session 2 plus
one tape from
Method
Temporal
correlations
No significant correlation
between alliance and subsequent
o
0
f Effects
0 0 1 1 0
o
1990 each of the change or prior change scores.
r
following
periods: weeks 4-
Feeley et
al., 1999
RCT data; Combined
sample: CT & CT +
MDD PHAS
(observer)
6; 7-9; 10-12
Same as
DeRubeis et al
Temporal
correlations
- p
No significant correlation
between alliance and subsequent
0a 0 0 1 1 0
Barber et
ADM (n = 25)
Combined data from Mixed CALPAS

1990
End of sessions Temporal

r e
change or prior change scores.
Alliance predicted symptom 0 0 1 1 1 0

al., 2000 four open trials;
Supportive-Expressive
Dynamic Therapy (n =
(patient) 2, 5, 10, and each
5th session
thereafter
correlations
l P
change. Depression predicted
late alliance but not early
alliance.
a
88)
n
Zuroff et RCT data; Combined MDD VTAS Sessions 3 & 15 Causal steps Increase in alliance partially 0a 0 1 0 1 0
r
al., 2000 sample: IPT, CBT, (observer) mediated the relationship
ADM + CM & placebo between perfectionism and
+ CM (n = 149)
u outcome.
Jo
Wilson et RCT data; Combined BN HRQ End of session 4 Causal steps Alliance did not mediate the 1 0 1 1 0 0
al., 2002 sample: CBT & IPT (n (patient) relationship between treatment
= 154) type and outcome.
Klein et RCT data; MDD WAI- Weeks 2 (after 3- Mixed effects Early alliance predicted change 1 0 1 1 1 0
al., 2003 CBASP (n = 169) Abbreviat 4 sessions), 6 growth in depressive symptoms, no
vs. ed 4-items (after 8-12 modeling treatment moderation. The
CBASP + ADM (n = (patient) sessions), and 12 reverse relationship was not
198) (after 16-20 observed.
sessions)
Baldwin Naturalistic database Mixed WAI Prior to session 4 Multilevel Therapist variability in alliance, 0 1 1 0 0 0
et al., collected from 45 (patient) mediation model but not patient variability,
2007 University counseling accounted for the relationship
centers; Various between pretreatment scores and
psychotherapy outcome.
treatment (n = 331)
Spinhove RCT data; BPD WAI-SF After 3, 15, and Longitudinal Alliance predicted change in 1 0 1 1 1 0
n et al., SFT (n = 44) (patient 33 months multilevel model symptoms. The reverse
2007 vs. and relationship was not observed.
TFP (n = 34) therapist)
Forbes et Academic treatment PTSD WAI-SF 3 weeks post Causal steps Alliance did not mediate the 0 0 1 1 0 0
al., 2008 center; Unspecified (patient intake relationship between anger at
psychotherapy and intake and post treatment
treatment (n = 103) therapist) symptoms
Byrd et Naturalistic database Mixed WAI-SF After session 3, Causal steps Alliance mediated the 0 0 1 0 0 0
f
al., 2010 from a University (patient) 4, or 5 with Sobel test relationship between comfort
training clinic; Various with closeness (attachment) and
psychotherapy
treatment (n = 66)
outcome
o
Owen et
al., 2011
Naturalistic database
from University
counseling center;
Mixed ITASr-SF
(patient)
End of academic
quarter
Causal steps
with bootstrap
method
Alliance mediated the
ro
relationship between clients'
p
perceptions of microaggressions
0 0 1 0 0 0
Crits-
Various psychotherapy
treatment (n = 232)
Data from a study on MDD CALPAS After sessions 3- Longitudinal -
and therapy outcomes
e
Alliance predicted next session 0 0 1 1 1 0
Christoph
et al.,
2011
training therapists;
Alliance-Fostering
Therapy (n = 45)
(patient) 16 multilevel model
r
symptom change. The reverse
relationship was observed only
P
in later treatment sessions.
Webb et
al., 2011
RCT data;
CT (n = 105)
MDD WAI-SF
(observer)
Session 3 and 3rd
to last session
a l
Multiple
regressions with
change scores
Early alliance significantly
predicted depressive symptom
improvement
0 0 1 0 1 0
Hirsh et
al., 2012
Subsample RCT data;
DBT (n = 43)
vs.
BPD WAI-SF
(patient)
Baseline, 4, 8,
and 12 months
u rnCausal steps
with product of
coefficient
relationship between
agreeableness and outcome in
1 0 1 0 1 0
o
GPM (n = 44) method DBT only
Strunk et
al., 2012
Falkenstr
om et al.,
RCT data;
CT + ADM arm (n =
176)
Primary Care;
MDD
Mixed
WAI-SF
(observer)
WAI-SF
(patient)
J Sessions 1-3
After every
session
Longitudinal
model
Longitudinal
multilevel model
Alliance scores did not predict
subsequent symptom change
Within-patient alliance predicted

next session symptom change
0a
0
0
1
1
1
1
1
1
1
0
2013 treatment including and vice versa. There was no

supportive, treatment moderation.
psychodynamic, and
CBT (n = 646)
Patterson University training Mixed WAI-SF After session 3 Causal steps Alliance did not mediate the 0 0 1 0 0 0
et al., clinic; Unspecified (patient) with Sobel test relationship between treatment
2013 psychotherapy (n = 68) expectancy and outcome
Sasso et RCT data; MDD WAI-SF Sessions 1-4 Longitudinal Within-patient and between- 0a 1 1 1 1 0
al., 2014 CT arm (n = 60) (observer) model patient alliance scores did not
predict subsequent symptom
change early in treatment
Yoo et al., Data from 13 Mixed WAI-SF After the 3rd Multilevel Alliance mediated the 0 0 1 0 0 0
2014 University counseling (patient) session mediation model relationship between treatment
centers, 4 community with bootstrap expectancy and outcome
counseling centers, and method
7 private practices;
treatment (n = 284)
Zilcha- RCT data; SET (n = MDD WAI Weeks 2, 4, 8, AR longitudinal Alliance predicted subsequent 1 0 1 1 1 0
Mano et 49) vs. CM + ADM (n (patient) and 16 multilevel model symptom levels, an effect not
al., 2014 = 51) vs. CM + PBO (n moderated by treatment type.
= 49)
McClinto
ck et al.,
2015
Naturalistic database
from University clinic;
Mixed WAI-SF
(patient)
Average of
Sessions 3-9
Causal steps
with bootstrap
method
relationship between expectancy
and outcome
o
0
f 0 1 1 0 0
ro
treatment (n = 116)
Burns et RCT data; Combined Chronic WAI-SF Week 4 and week Cross-lagged Alliance was associated with 0a 0 1 1 1 0
al., 2015 sample:
Enhanced CBT & CBT
(n = 94)
Pain (patient) 8 panel
correlations
p
subsequent symptom change.
-
Xu et al.,
2015
Naturalistic sample
from university
Mixed WAI-SF
(patient)
Prior to each
session starting
Latent change
score modeling
r e
Alliance predicted subsequent
symptom improvement and vice
0 0 1 0 1 0
training clinic; Various

psychotherapy
treatment (n = 638)
with session 3
l P
versa
Klug et
al., 2016
Data from comparative
trial in outpatient
MDD HAQ
(patient
Every 3 months
(CBT) and 6
n a
Multilevel
mediation model
Alliance did not mediate the
relationship between treatment
1 0 0 1 0 0
r
university clinic; and months (PA, PD) type and outcome
PA (n = 35) therapist)
vs.
PD (n = 31)
u
Jo
vs.
CBT (n = 34)
Kushner RCT data; MDD CALPAS 3rd and 12th Serial multiple Alliance mediated the 1 0 1 0 1 0
et al., ADM (n = 74) (patient session mediation model relationship between
2016 vs. and with bootstrap agreeableness and symptom
IPT (n = 65) therapist) method change, an effect not moderated
vs. by treatment
CBT (n = 70)
Maitland RCT data; Mixed WAI-SF Average of Causal steps Alliance mediated the 1 0 0 0 0 0
et al., FAP (n = 11) (patient) sessions 1-3 with bootstrap relationship between treatment
2016 vs. method condition symptom change.
WW (n = 11)
Falkenstr RCT data from MDD WAI-SR After every Dynamic panel Alliance predicted next session 1 1 1 1 1 0
om et al., community-based (patient) & session data model symptom change. Results were
2016 psychiatric clinic; WAI-SF not moderated by treatment and
CBT (n = 43) (therapist) the reverse relationship was not
vs. observed.
IPT (n = 41)
Zilcha- RCT data; Mixed WAI Sessions 1-4 Multilevel Early alliance development 0a 0 1 0 1 0b
Mano et Various psychotherapy (patient) mediation model predicted treatment outcome for
al., 2017 treatment (n = 166) patients with pretreatment
with therapists interpersonal problems
randomized to different
feedback conditions
Lawson et University training PTSD ITA-RS Session 3 or 4 Causal steps Alliance mediated the 0 0 1 0 0 0
al., 2017 clinic; Integrated (patient) with bootstrap relationship between baseline
Relationship and method interpersonal problems and
Trauma-Based CBT (n dissociation posttreatment, but
= 76) not between interpersonal
problems and trauma symptoms
posttreatment
o f
Renner et
al., 2018
Specialized care
facility; SFT (n = 20)
MDD SRS
(patient)
After every
session
AR longitudinal
multilevel model
r o
Alliance did not predict change
in depressive symptoms nor vice
versa.
0 0 0 1 1 0
Sauer-
Zavala et
RCT data;
UP (n = 77)
Mixed
Anxiety
WAI-SF
(patient)
After session 4 Causal steps
with bootstrap
- p
relationship between treatment
1 0 1 0 0 0
al., 2018 vs.

SDP (n = 76)
Disorders method
e
expectancy and change in
r
symptoms in SDP but not UP.
Santoft et
al., 2019
RCT data; CBT (n =
40)
vs.
Exhaustio
n Disorder
WAI-SF
(patient)
After every
session
Multilevel
mediation model
l P
Therapeutic alliance did not
mediate the relationship between
condition and burnout.
1 0 1 1 1 0
a
RTW-I (n = 42)
n
Rubel et RCT data; CBT (n = GAD WAI-SF After every Dynamic Within-patient alliance scores 1 1 0 0 1 0
r
al., 2019 57) with patients (patient) session structural were associated with reduction in
randomized to three equation anxiety and increase in coping
different priming
conditions (n = 19 per
u modeling experiences during the following
session. Results were not
Jo
condition) moderated by condition.
Brattland Naturalistic RCT; Mixed WAI-SF At Session 1 and Multilevel Alliance mediated the 1 0 1 0 0 0
et al., various psychotherapy (patient) after 2 months of mediation model relationship between treatment
2019 treatment with patients treatment condition and outcome.
randomized to TAU (n
= 74) or ROM (n = 69)
Gomez- RCT data; EBCT (n = MDD WAI-SF After every Hybrid random Within-patient and between- 1 1 1 0 1 0
Penedo et 70) vs. CBT (n = 71) (patient) session effects model patient alliance predicted next
al., 2020 session symptomatology even
when adjusting for treatment
condition
Leibovich RCT data; combined MDD WAI-SF Session 4 Causal steps Alliance mediated the 0a 0 1 0 0 0
et al., sample of supportive (patient) with bootstrap relationship between supportive
2020 therapy vs supportive- method techniques and outcome
expressive therapy (n =
61)
Sullivan Naturalistic database Mixed ITA-RS Early (session 3 Causal steps Alliance did not mediate the 0 0 1 1 0 0
et al., from a University or 4); middle with bootstrap relationship between early
2020 training clinic; trauma- (between sessions method interpersonal distress and
based CBT and 6-8); late outcome
relational-based CBT (between sessions
(n = 137) 16-24)
Note: Column Headings: RCT = Randomized Controlled Trial; n 40 = Sample size per treatment arm is at least 40 or, combined is at least 40 if study did not examine treatment
effects; Control = Control Group; Multiple Mediators = Study included more than alliance as a potential mediator; Temporality = Study included two or more assessments of
alliance during treatment phase; Manipulation = Manipulation of Alliance; 0 = Absent/No; 1 = Present/Yes. *Denotes reported sample size used in analysis; a Denotes data comes
from RCTs but authors do not make use of RCT design in analyses such as by looking at treatment moderation; b Study randomized clinicians to different kind of feedback
pertaining to the alliance but did not examine this manipulation in the analyses. Interventions: CT = Cognitive Therapy; ADM = Antidepressant Medication; IPT = Interpersonal
Therapy; CBT = Cognitive Behavioral Therapy; EBCT = Exposure-Based Cognitive Therapy; CM = Clinical Management; CBASP = Cognitive Behavioral Analysis System of
f
Psychotherapy; SFT = Schema Focused Therapy; TFP = Transference Focused Psychotherapy; PA = Psychoanalytic; PD = Psychodynamic; DBT = Dialectical Behavior Therapy;
GPM = General Psychiatric Management; FAP = Functional Analytic Psychotherapy; WW = Watchful Waiting; UP = Unified Protocol; SDP = Single Disorder Protocols
o
(empirically supported); RTW-I = Return to Work Intervention; ROM = Routine Outcome Monitoring. Diagnosis: MDD = Major Depressive Disorder; BN = Bulimia Nervosa;
r o
BPD = Borderline Personality Disorder; PTSD = Posttraumatic Stress Disorder; GAD = Generalized Anxiety Disorder. Alliance Measures: WAI = Working Alliance Inventory;
WAI-SF = Working Alliance Inventory Short Form; WAI-SR = Working Alliance Inventory Scale Revised; CALPAS = California Psychotherapy Alliance Scale; PHAS = Penn
- p
Helping Alliance Scale; VTAS = Vanderbilt Therapeutic Alliance Scale; HRQ = Helping Relationship Questionnaire; HAQ = Helping Alliance Questionnaire; HAQ-R = Helping
Alliance Questionnaire-Revised; HAq-II = Helping Alliance Questionnaire-II; ITASr-SF = Individual Treatment Alliance Scale Revised-Short Form; ITA-RS = Individual
e
Therapy Alliance Revised/Shortened; SRS = Session Rating Scale. AR = autoregressive.
P r
a l
r n
u
Jo
CHANGE Pre-proof 42
Table 2.
Number (%) of studies meeting criteria for process research (n = 37)
Requirement n studies (%)
RCT, yes, n (%) 13 (35.1)
Disaggregated within and between-patient effects, yes, n (%) 7 (18.9)
Sample size per condition 40, yes, n (%) 30 (81.1)
Multiple mediators, yes, n (%) 19 (51.4)
of
Temporality, yes, n (%) 22 (59.5)
ro
Manipulation of mediator, yes, n (%) 0 (0.00)
-p
Note. RCT = randomized controlled trial; n = number of studies, % = percent of studies
re
lP
na
ur
Jo
Journal Pre-proof
Running head: ALLIANCE AS A MEDIATOR OF CHANGE 43
Figure 1.
PRISMA flow diagram of study identification and selection process
Unique records identified Additional records identified

through PsycINFO database through reference lists of other
Identification
searching with initial screening reviews, meta-analyses, and

criteria manuscripts
(n = 1,613) (n = 59)
of
Records after duplicates removed and
screened on the basis of title (n = 1,651)
ro
Records excluded (n = 287):
-p
Screening
- No mediation (n = 28)
- Case study (n = 38)
Records screened on basis
- Group treatment (n = 7)
re
of abstract
- Not adult (n = 14)
(n = 479)
- Not empirical (n = 137)
lP
- Not psychotherapy (n = 59)

- Telehealth (n = 4)
na
Eligibility
Full-text articles excluded (n = 155):

Full-text articles assessed - No mediation (n = 77)
ur
for eligibility - Case study (n = 1)

(n = 192) - Group treatment (n = 13)
- Inpatient treatment (n = 4)
Jo
- No validated alliance measure (n =

5)
- Not adult (n = 3)
- No validated symptom outcome
measure (n = 13)
Included
- Not empirical (n = 17)

Included
- Not psychotherapy (n = 4)
Studies included in present - Telehealth (n = 8)
review - Same data as another study (n =
(n = 37) 10)
CHANGE Pre-proof 3
Role of funding sources

This work was supported by the National Institute of Mental Health (R01 MH066348) awarded
to NCF. The funding source had no role in study design, analysis, writing or decision to submit
this paper for publication.
Contributors
ALB developed the concept for the study and reviewed the literature. ACK and NCF consulted
on scope of the review and relevant methodology such as search terms and inclusion/exclusion
criteria. ALB wrote the first draft of the manuscript and all three authors contributed to and have
approved the final manuscript.
Conflict of interest
All authors declare that they have no conflicts of interest.
of
Acknowledgments
The authors wish to thank Alexandra Bowling for her diligent review of articles included in this
ro
review.
-p
re
lP
na
ur
Jo
CHANGE Pre-proof 4
Highlights
 We evaluated studies examining the alliance as a mediator of change in psychotherapy.
 Alliance plays an important role pantheoretically.
 Studies displayed methodological and statistical heterogeneity.
 Steps for improving future mediation research are proposed.
of
ro
-p
re
lP
na
ur
Jo
CHANGE Pre-proof 5
Author Biographies
Allison L. Baier, M.A., is a doctoral candidate at Case Western Reserve University. Her research
interests include understanding mechanisms underlying treatment outcomes and effectively
increasing dissemination and implementation of evidence-based interventions for PTSD.
Alexander C. Kline, Ph.D., is a graduate of Case Western Reserve University and current
Postdoctoral Fellow at UCSD/VA San Diego Healthcare System. His research focuses on
interventions for PTSD and related comorbidities, with emphasis on processes and predictors
linked to clinical outcomes.
Norah C. Feeny, Ph.D., is a Professor in the Department of Psychological Sciences at Case

Western Reserve University. Her research interests include examining patient preferences,
of
evaluating interventions for PTSD, and understanding what predicts who will benefit from these
treatments.
ro
-p
re
lP
na
ur
Jo

Therapeutic Alliance As A Mediator of Change A Systematic Review and Evaluation of Research (Baier, Kline & Feeny, 2020)

Uploaded by

Copyright:

Available Formats

Therapeutic Alliance As A Mediator of Change A Systematic Review and Evaluation of Research (Baier, Kline & Feeny, 2020)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Therapeutic Alliance As A Mediator of Change A Systematic Review and Evaluation of Research (Baier, Kline & Feeny, 2020)

Uploaded by

Copyright:

Available Formats

Journal Pre-proof

Therapeutic alliance as a mediator of change: A systematic review

Allison L. Baier, Alexander C. Kline, Norah C. Feeny

To appear in: Clinical Psychology Review

Received date: 25 November 2019

© 2019 Published by Elsevier.

Therapeutic Alliance as a Mediator of Change: A Systematic Review and Evaluation of Research

Allison L. Baiera, Alexander C. Klinea1, & Norah C. Feenya

Psychological Sciences, 11220 Bellflower Road, Cleveland, OH, USA, 44106-7123

Corresponding author: Allison L. Baier, Case Western Reserve University, Department of

Psychological Sciences, 11220 Bellflower Road, Cleveland, Ohio 44106. Email:

alb184@case.edu; Phone: 216-368-8934.

mechanism underlying psychotherapy response. This systematic review evaluated research

examining the alliance as a potential mediator of symptom change, reviewing study

therapeutic outcomes in 70.3% of the studies. We observed significant heterogeneity across

studies in terms of methodology, including timing of alliance assessment, study design,

constructs used in mediation models, and analytic approaches. Building on recent

Keywords: alliance, mediator, mechanism, psychotherapy, systematic review

Therapeutic Alliance as a Mediator of Change: A Systematic Review and Evaluation of Research

psychotherapies, including an increasing number of evidence-based psychotherapies with well-

response or nonresponse, thereby optimizing treatments and ultimately providing more

personalized, effective care.

One hypothesized process underlying treatment response is the therapeutic alliance.

associated with positive treatment outcomes across a range of psychotherapies as evidenced by

.28; k = 190; Martin, Garske, & Davis, 2000: r = .22, k = 79).

2015). As a consequence, some methodologists argue alliance is a nonspecific treatment factor

process impacts outcomes. Additionally, a greater understanding of the alliance’s role in

which treatments lead to change.

Requirements for a Mediator

mediation analyses (Kazdin, 2007).

A mediator (variable M) is an intervening variable that statistically explains or accounts

methods, readers are directed to MacKinnon and colleagues’ review (2007).

In addition to the mechanics of the mediation analysis, methodologists are increasingly

advocating for researchers to disentangle within-patient and between-patient effects in process

between-patient variation in alliance scores, is an important consideration for process research

that seeks to uncover why and how change occurs.

In sum, methodological approaches are integral to identifying and assessing proposed

by the types of analyses employed in studying this process.

Advances in Assessing Mediation

Mediators are an important first step in understanding hypothesized change processes;

hypothesized mediator through experimental designs (termed, ‘experimental manipulation’) to

mechanisms of change, some experts suggest that in statistical mediation, temporality,

statistical methodology) to better understand mediators in psychotherapy.

To date, meta-analyses have examined the alliance-outcome relationship as well as

alliance serves as a potential mediator of change in psychotherapy—and whether alliance may be

understanding the specific change processes occurring in treatment.

Aims of Current Review

potential role of the therapeutic alliance as a mediator of change in psychotherapy. As therapy

alliance is supported as a possible mediator of therapeutic change.

treatment(s)/intervention(s)” OR “Interpersonal (psycho)therapy” OR “psychodynamic/analytic

“mechanism” OR “mechanisms of change/action” OR “working mechanisms (of

psychotherapy)” OR “processes of therapy” OR “process research” OR “change” OR

commentaries, theoretical essays, meta-analyses, systematic reviews, and qualitative studies

characteristics included: psychotherapy intervention type(s), primary diagnosis of the sample,

one study, assessment of temporality (defined as 2 or more assessments of the therapeutic

average), and direct experimental manipulation of alliance. We chose to follow guidelines

1), Psychodynamic Therapy (PD, k = 1), Cognitive Behavioral Analysis System of

Psychotherapy (CBASP, k = 1), Transference Focused Psychotherapy (TFP, k = 1), Functional

(RTW-I, k = 1). Six studies included a combined treatment or pharmacologic treatment.