UvA-DARE (Digital Academic Repository)

Mechanisms of change in psychotherapy for depression

An empirical update and evaluation of research aimed at identifying psychological mediators
Lemmens, L.H.J.M.; Müller, V.N.L.S.; Arntz, A.; Huibers, M.J.H.
DOI
10.1016/j.cpr.2016.09.004
Publication date
2016
Document Version
Final published version
Published in
Clinical Psychology Review
License
Article 25fa Dutch Copyright Act (https://www.openaccess.nl/en/in-the-netherlands/you-share-
we-take-care)
Link to publication

Citation for published version (APA):

Lemmens, L. H. J. M., Müller, V. N. L. S., Arntz, A., & Huibers, M. J. H. (2016). Mechanisms
of change in psychotherapy for depression: An empirical update and evaluation of research
aimed at identifying psychological mediators. Clinical Psychology Review, 50, 95-107.
https://doi.org/10.1016/j.cpr.2016.09.004

General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s)
and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open
content license (like Creative Commons).

Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please
let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material
inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter
to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You
will be contacted as soon as possible.

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

Download date:03 nov. 2023

 Clinical Psychology Review 50 (2016) 95–107

Contents lists available at ScienceDirect

Clinical Psychology Review

journal homepage: www.elsevier.com/locate/clinpsychrev

Review

Mechanisms of change in psychotherapy for depression: An empirical

update and evaluation of research aimed at identifying
psychological mediators
Lotte H.J.M. Lemmens, PhD a,⁎, Viola N.L.S. Müller, MSc b,
Arnoud Arntz, Prof PhD a,c, Marcus J.H. Huibers, Prof PhD a,d
a
Department of Clinical Psychological Science, Faculty of Psychology and Neuroscience, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlands
b
Department of Psychology, University of Trier, Am Wissenschaftspark 25-27, 54286 Trier, Germany
c
Department of Clinical Psychology, University of Amsterdam, PO Box 19268, 1000 GG Amsterdam, The Netherlands
d
Department of Clinical Psychology, VU University Amsterdam, Van der Boechorststraast 1, 1081 BT Amsterdam, The Netherlands

H I G H L I G H T S

• Little is known about the psychological mechanisms of psychotherapy for depression.

• The mechanism question has motivated dozens of investigations of mediation.
• We provide an empirical update and critical evaluation of this body of research.
• Research is heterogeneous and unsatisfactory in methodological respect.
• Psychotherapy might be too complex to be explained in simple models of psychological change.

a r t i c l e i n f o a b s t r a c t

Article history: We present a systematic empirical update and critical evaluation of the current status of research aimed at iden-
Received 8 June 2015 tifying a variety of psychological mediators in various forms of psychotherapy for depression. We summarize
Received in revised form 16 September 2016 study characteristics and results of 35 relevant studies, and discuss the extent to which these studies meet several
Accepted 18 September 2016
important requirements for mechanism research. Our review indicates that in spite of increased attention for the
Available online 20 September 2016
topic, advances in theoretical consensus about necessities for mechanism research, and sophistication of study
Keywords:
designs, research in this field is still heterogeneous and unsatisfactory in methodological respect. Probably the
Mediators biggest challenge in the field is demonstrating the causal relation between change in the mediator and change
Mechanisms in depressive symptoms. The field would benefit from a further refinement of research methods to identify
Psychotherapy processes of therapeutic change. Recommendations for future research are discussed. However, even in the
Depression most optimal research designs, explaining psychotherapeutic change remains a challenge. Psychotherapy is a
multi-dimensional phenomenon that might work through interplay of multiple mechanisms at several levels.
As a result, it might be too complex to be explained in relatively simple causal models of psychological change.
© 2016 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
1.1. Requirements for a mediator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
1.1.1. Statistical mediation is important but not sufficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
1.1.2. Requirements for study designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
1.2. Research studying mediators in psychotherapy for depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
1.3. Aim of the current review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

⁎ Corresponding author.
E-mail address: [email protected] (L.H.J.M. Lemmens).

http://dx.doi.org/10.1016/j.cpr.2016.09.004
0272-7358/© 2016 Elsevier Ltd. All rights reserved.
96 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107

2. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.1. Data sources and data reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.2. Data assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.1. Study characteristics and results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.2. A closer look at the value of these results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.2.1. The criteria in concert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Appendix A. Key-term scheme for database search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
104
Appendix B. In & exclusion criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
105
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
106

1. Introduction MacKinnon, Lockwood, & Williams, 2004; Shrout & Bolger, 2002). The
applicability of the model in this field is further limited by restrictions
Many researchers in the field of clinical psychology agree that resulting from the first and fourth criterion. The first criterion (efficacy
gaining a better understanding of the mechanisms underlying psycho- test) is formulated in a way that the ability to perform mediation anal-
therapeutic change is crucial for optimizing treatment outcomes for pa- ysis strongly depends on the presence of differential treatment effects.
tients suffering from psychiatric disorders such as depression (Kazdin & When two treatments turn out to be equally effective – a phenomenon
Nock, 2003; Kraemer, Wilson, Fairburn, & Agras, 2002). Knowledge that is not uncommon in the field of psychotherapy for depression (for
about active ingredients of therapy can assist in the verification and re- more details see e.g. Cuijpers & van Straten, 2011; Cuijpers, van Straten,
finement of theories of the disorder, and allows enhancement of ele- Andersson, & van Oppen, 2008; Wampold et al., 1997) – this type of me-
ments that are crucial for therapeutic change, while dismissing those diation analysis is not possible. This is an important drawback, because
found to be redundant (Garratt, Ingram, Rand, & Sawalani, 2007; especially when two treatments turn out to be equally effective it is im-
Longmore & Worrell, 2007). portant to examine processes of change, since this can tell us more
An important first step towards examination of mechanisms of about whether the change that is observed is reached through similar
change is the identification of mediators (Kazdin & Nock, 2003; or differential pathways (MacKinnon, 2008). Moreover, given the popu-
Kraemer, Stice, Kazdin, Offord, & Kupfer, 2001; Kraemer et al., 2002). lation (depressed patients) and the nature of treatments (psychothera-
A mediator is a variable that statistically explains why and in what py), it is ethically and practically very difficult (if not impossible) to
way a treatment has an effect on outcome, and can be seen as a potential include a substantially less powerful treatment (such as a full waiting-
mechanism: the actual process or event that is responsible for change list control group, or a placebo intervention) to increase the contrasts
(Baron & Kenny, 1986; Kazdin, 2007, 2009; Kraemer et al., 2001; between groups. And even if a third ineffective control condition
MacKinnon, Fairchild, & Fritz, 2007). In other words, the mechanism is would be added, it is still not possible to test differential pathways be-
the phenomenon to reveal, the mediator can be the mean to this end. tween the two equally effective treatments. The fourth Baron and
Mediators can be distinguished from moderators in the sense that Kenny (1986) criterion (mediation test) has been criticised because
they explain the relationship between an independent and dependent the tests that have to demonstrate the reduction of the effect after sta-
variable (i.e. they indicate whether treatment has an effect on outcome tistically controlling for the mediator have shown to be underpowered
via the mediator), whereas moderators influence that relationship (i.e. (MacKinnon et al., 2007).
they indicate when or under what conditions the relationship between As a result of these limitations, the criteria for statistical mediation
treatment and outcome can be expected: Hayes, 2013). have been modified over time to make them more applicable and suit-
able for treatment research. For example, the MacArthur group
1.1. Requirements for a mediator (Kraemer et al., 2001, 2002) toned down the importance of the first cri-
terion by stating that differential treatment effects are not required to
Establishing a mediator involves several requirements. For a long establish mediation as long as there is an interaction between treatment
time, mediation solely referred to statistical mediation: to statistically and the mediator. This is particularly useful in clinical trials comparing
demonstrate that the effect of treatment on outcome is explained by a two (equally) effective treatments that are likely to operate through dif-
third variable: the mediator. The most well-known method to deter- ferent mechanisms. With regard to step 4, it was decided that it was suf-
mine statistical mediation is indubitably Baron and Kenny's (1986) ficient to show that treatment has an effect on the mediator and that the
causal step method. With almost 60.000 citations, their paper is one of mediator has an effect on the outcome, even after controlling for treat-
the most frequently cited articles in the field of psychology. According ment, a procedure known as joint significance testing (MacKinnon et al.,
to Baron & Kenny, mediation is established when 1) there is a main ef- 2007). Furthermore, advances have been made in statistical methods to
fect of treatment (efficacy test), 2) treatment is related to change in the test the various mediation models (see developments by e.g. Arbuckle,
mediator (intervention test), 3) change in the mediator and change in 1999, 2005; Kraemer et al., 2001, 2002; MacKinnon et al., 2002, 2004,
outcome are related (psychopathology test), and 4) the effect of treat- MacKinnon et al., 2007, MacKinnon, 2008: Muthén & Muthén, 2001,
ment on outcome is absent (full mediation) or significantly weakened 2007; Preacher & Hayes, 2004).
(partial mediation) when statistically controlling for the mediator
(mediation test). Subsequently, a Sobel test (Sobel, 1982) determines 1.1.1. Statistical mediation is important but not sufficient
the amount of mediation – also called the indirect effect. Although statistical mediation still plays a central role in addressing
Influential as it has been, the Baron and Kenny (1986) model has sig- whether a particular construct accounts for change (Hollon & DeRubeis,
nificant limitations for application in social sciences and therefore also 2009; Kazdin, 2007, 2009), it is not sufficient to make a case for the op-
in clinical process research for disorders such as depression. For exam- eration of a mediator (e.g. Johansson & Høglend, 2007; Kazdin, 2007,
ple, the method has low type I error rates and, in order to have sufficient 2009; Laurenceau, Hayes, & Feldman, 2007). Probably the most impor-
power, requires large sample sizes and large treatment effects, both of tant addition to statistical mediation is demonstrating the direction of
which are not always available in this type of research (Hoyle & causality. Conditions for inferring causal relations in scientific research
Kenny, 1999; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002; have been outlined by e.g. Hill (1965), Kenny (1979), Schlesselman
 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107 97

(1982), and brought to the psychotherapy literature by Kazdin (2003, disorders. A closer look at the literature specific for depression indicates
2007, 2009). Apart from a strong statistical association between treat- that the majority of studies has focused on the mediational role of cog-
ment, mediator and outcome, Kazdin describes six requirements for ad- nitive processes, such as automatic thoughts, dysfunctional attitudes,
equate evidence for causal temporal relationships. First of all, it has to be attributional style, and other cognitive distortions. The cognitive medi-
demonstrated that the treatment causes the mediator variable to ation hypothesis was also the focus of the influential systematic review
change, which in turn causes the outcome, and not the other way by Garratt et al. (2007). Garratt and colleagues summarized results of 31
around (Kazdin & Nock, 2003; Kraemer et al., 2002). In order to get a studies on the role of cognitive change and concluded that research gen-
clear view of the shape of change and the relation between mediator erally supports the cognitive mediation hypothesis, but that this does
and outcome, it is important that both the mediator and outcome mea- not necessarily need to be specific for interventions in which cognitions
sure are assessed at multiple time points during treatment. The impor- are actively targeted. This indicates that cognitive change, no matter
tance of demonstrating temporality is supported by many research how it occurs, might play a role in various treatment modalities. Even
groups (e.g. Collins & Graham, 2002; Hollon & DeRubeis, 2009; though Garratt et al. acknowledged that these findings increased
Johansson & Høglend, 2007; Kazdin, 2007, 2009; Kazdin & Nock, 2003; knowledge about the relation between cognition and depression, they
Kraemer et al., 2002; Laurenceau et al., 2007; Murphy, Cooper, Hollon, emphasized that their findings did not permit clear-cut answers about
& Fairburn, 2009), and has even been called the fifth step of statistical the exact role of cognitive change as a process that facilitates psycho-
mediation analysis (Johansson et al., 2010). Second, alternative explana- therapeutic change in the context of psychotherapy. They provided sev-
tions for the observed relation between mediator and outcome should eral reasons for this. First of all, there was a large variety in research
be ruled out. This can be done by using an experimental approach in questions and methodology across studies, which made it difficult to
which all variables are held constant across individuals in various condi- compare results across studies and to integrate findings into broader
tions while changing only the proposed mechanism of change (Kazdin, knowledge. Second, many studies did not meet the criteria for reputable
2007, 2009). Furthermore, Kazdin emphasizes the importance of speci- mechanism research, hereby limiting the interpretability of study find-
ficity of the association among the intervention, proposed mediator ings. More specifically, Garratt et al. concluded that none of the studies
and outcome. This means that it has to be demonstrated that the medi- that were identified in their review addressed the criteria for mediation
ator plays a crucial role in one treatment, but not (or less so) in the in methodologically sound ways. Garratt and colleagues expressed their
other. In addition, inclusion of plausible processes, consistency across hope that this would change in subsequent years, in studies with e.g.
studies, and a gradient, in which larger changes in the mediator are as- larger sample sizes, up-to-date-statistical methods, and a broader
sociated with larger changes in outcome, should further enhance the array of measures. These issues are acknowledged by others in the
evidence. field as well (e.g. Johansson & Høglend, 2007; Kazdin, 2007; Kraemer
Kazdin (2007) emphasizes that each criterion is important, but that et al., 2001; Laurenceau et al., 2007). A third difficulty in interpreting re-
interpretations should be made based on their convergence. Examina- sults from studies in this field – not mentioned by Garratt et al. – is the
tion starts with statistical tests for mediation. After that, one determines fact that not every study that makes claims about mediators, actually
the value of the results by examining the extent to which a study meets performed statistical mediation analyses. Instead, some studies present
the other criteria. Even though the satisfaction of each criterion in- correlations between changes in hypothesized process measures and
creases the strength of the argument for the operation of a mediator – depressive symptoms from pre- to post-treatment as evidence for me-
or even a mechanism – not all criteria are weighted equally important. diation. Others make claims about mediators based on prediction anal-
According to Kazdin and Nock (2003), statistical association, temporality, yses. This does not only further increase the heterogeneity in the field,
specificity, and experiment are considered to be the most important, but also leads to conclusions about mediators in studies where no statis-
whereas the remaining three should further enhance the evidence. tical mediation analyses were performed. Garratt and colleagues did not
differentiate between this in their review. Fourth, since most studies so
1.1.2. Requirements for study designs far mainly focused on the role of cognitive factors, the influence of non-
cognitive factors is still largely unknown.
The extended requirements and possibilities for identifying mediators
also called for additional features of study designs. According to the latest 1.3. Aim of the current review
standards, the extent to which a process meets the requirements for me-
diation can only be examined properly in a theoretically well planned RCT Almost ten years have passed since the Garratt et al. (2007) review,
with carefully spaced repeated measures, sufficient power and an appro- and the question is whether and how the field has changed. The aim of
priate control group (Kazdin, 2007; Kazdin & Nock, 2003; Kraemer et al., the current review was therefore to provide an update and critical meth-
2002; Laurenceau et al., 2007). Furthermore, it is important to experimen- odological evaluation of the current body of research on this topic. In a
tally manipulate the proposed mediators, which requires an experimental systematic literature search, we selected studies aimed at identifying psy-
study design. In addition, mediation analysis should be performed using chological mediators in psychotherapy for depression. To get a compre-
up-to-date definitions and state-of-the-art statistical analyses techniques hensive overview of the field, we included various forms of
(Collins & Graham, 2002; Haaga & Stiles, 2000; Haubert & Dobson, 2007; psychotherapy and included both cognitive and non-cognitive processes.
Kraemer et al., 2002; Laurenceau et al., 2007; MacKinnon et al., 2007). We only selected studies that included an actual test of statistical media-
Moreover, depending on what the theory stipulates about processes, as- tion (Baron & Kenny (1986) or one of the more advanced methods). We
sessment of a single mediator might not be sufficient. It is therefore rec- summarize study characteristics and results of 35 studies and discuss the
ommended to include multiple mediators to examine rival hypotheses, extent to which these studies meet the most important requirements for
test alternative explanatory models, and map out interactions between mechanism research that were mentioned earlier. With this we hope to
theorized processes. learn more about the magnitude and relevance of the existing body of re-
search and map out necessities for future research.
1.2. Research studying mediators in psychotherapy for depression
2. Method
The past decades, the interest for mediators in mechanism research
in depression has grown, and several research groups worldwide have 2.1. Data sources and data reduction
studied mediators of psychotherapy. In 2007, Johansson and Høglend
identified 61 studies that performed mediational analyses to identify Three different approaches were used to identify relevant studies.
the active ingredients of psychotherapy for several psychiatric First, five databases (i.e. PubMed, PsychInfo, Embase, Cochrane, and
98 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107

Cinahl) were systematically searched for potentially relevant papers goal in process research is to identify any factors that facilitate symptom
that were published in English in peer reviewed journals until spring change, regardless of their specificity to one treatment. A first priority is
2016. Key terms were Depression, Psychotherapy, Mechanisms and Me- therefore to identify process factors that are a linking pin between treat-
diation (a full key-term scheme can be found in Appendix A). The data ment and outcome. A subsequent specificity analysis could then show
search yielded a total of 617 unique studies. One of us (VM) carefully whether this factor plays a role in only this treatment or also in other
read through all abstracts1 and retained those articles that met a set of treatments. By requiring specificity as a (testable) criterion for media-
a priori generated inclusion and exclusion criteria. LL checked the gen- tion, basic information about whether or not a process facilitates symp-
erated table entries for accuracy. tom change is discarded when it turns out that the specificity criterion is
To be included in the review articles needed to be empirical research not met. Since we consider this information important, we decided not
reports (no reviews, theoretical essays or commentaries) examining to include specificity in our evaluation.
psychological mediators over the course of treatment of various forms
of evidence-based psychotherapy for patients (adults and adolescents)
with (subclinical) depression. Furthermore, studies needed to actually 3. Results
include statistical mediation analyses in their analysis plan (in the
sense of Baron and Kenny or one of the modern alternatives). Studies 3.1. Study characteristics and results
including patients diagnosed with bipolar depression were excluded,
as were those that focused on other forms of psychopathology2 and/or Table 1 (left panel) gives an overview of study characteristics and re-
(relapse) prevention. A complete overview of the in- and exclusion sults of 35 studies that were included in the review. The majority of
criteria can be found in Appendix B. studies was conducted in the USA (57.1% vs. 28.6% in Europe, and
Of the 617 articles that were identified in the literature search, 584 14.3% in other parts of the world), and 48.6% was published in the
did not meet our inclusion criteria and were therefore excluded. The past five years (2012–2016). Sample sizes ranged between n = 4 and
majority of studies were excluded because they did not focus on psy- n = 523, with a mean of n = 173 (SD = 145.3). Patients were adults
chological mechanisms of treatment for depression (n = 356). Other (in 26 studies) and adolescents (in 9 studies) ranging in age from 12
papers were excluded because they were theoretical papers (e.g. re- to 68 years (M = 40.2 SD = 8.2 for studies in adults4 and M = 15.1
views, commentaries) instead of empirical research reports (n = 90), SD = 0.5 for studies including adolescents5). In 90.9% of the studies
or because they did not focus on an (evidence-based) psychological the majority (N 50%) of participants were female.6
intervention (n = 135). Two papers were excluded because after Cognitive (Behavioural) Therapy (C(B)T) was the most frequently
careful reading they did not perform statistical mediation analysis researched intervention (examined in 21/35 studies), followed by Mind-
(Backenstrass et al., 2006; Sasso, Strunk, Braun, DeRubeis, & Brotman, fulness Based Cognitive Therapy (MBCT, included in 5 studies). Other
2015), and one because it did not include a clinical outcome measure treatments were Acceptance and Commitment Therapy (ACT, k = 3), Be-
(Johansson et al., 2010).3 A total of 33 articles met all inclusion criteria havioural Activation (BA, k = 1), Cognitive Behavioural Analysis System
and were selected for further review. Subsequently, we hand searched of Psychotherapy (CBASP, k = 1), Interpersonal Psychotherapy (IPT,
reference lists of the 33 articles that met all inclusion criteria, and k = 2), Non-Directive Supportive Therapy (NST, k = 2), Problem Solving
asked several experts (3 psychologists, 1 psychiatrist) with (Couples) Therapy (PST, k = 2), Psychodynamic Therapy (k = 1), Psycho-
longstanding experience in the research field and clinical practice of de- analytic Therapy (k = 1), and Systematic Behavioural Family Therapy
pression to check the list that was generated. Two additional papers (SBFT, k = 2). Three studies included a combined treatment.
were added, resulting in a total of 35 studies that were further explored. Common measures of depression severity were the (second edition
of the) Beck Depression Inventory (BDI(-II); Beck, Steer, & Brown, 1996;
Beck, Ward, Meldelson, Mock, & Erbauch, 1961), which was implement-
2.2. Data assessment ed in 18 studies, and the Hamilton Rating Scale for Depression (HRSD;
Hamilton, 1960), used in 7 studies. Nine studies (e.g. van Aalderen et
Two researchers (LL and VM) carefully read the 35 articles that were al., 2012; DeRubeis et al., 1990; Vittengl, Clark, Thase, & Jarrett, 2014;
selected and tabulated study characteristics and results. To answer our Warmerdam, van Straten, Jongsma, Twisk, and Cuijpers, 2010) used
main research question, all papers were assessed by means of several them both, thereby obtaining a self-report and an observer-based mea-
important requirements for mediation research that were discussed sure of depression.
earlier: the use of an RCT design and inclusion of a control group, a suf- The identified studies examined 39 different potential mechanisms.
ficient sample size (defined as n ≥ 40), examination of multiple potential Given the substantial number of studies that examined C(B)T, media-
mediators within one study, the assessment of temporality (as defined tors were predominantly the theorized processes of this intervention,
by 3 or more assessments in the treatment phase), and direct experi- such as Negative (Automatic) Thoughts (7 studies), Dysfunctional Atti-
mental manipulation of the mediator. Each study was rated with respect tudes (7 studies), Attributional style (3 studies) and other cognitive
to meeting (+) or not meeting (−) each of these criteria. Differences in constructs (9 constructs in 7 studies). Furthermore, six studies assessed
scoring were resolved by consensus. A qualitative analysis was conduct- the behavioural component of CBT. In studies in which Mindfulness-
ed by summarizing, comparing and contrasting the data. Based interventions were the choice of treatment Rumination, Mindful-
It has to be noted that specificity is not included in the list of features ness, and Worry were common process measures (included in 5, 4, and
that was described above. This does not mean that we think that exam- 3 studies respectively). The potential mediational role of Therapeutic Al-
ining specificity is not important (in fact, as was stated in the liance was examined in 3 of the 35 identified studies. As can be seen in
Introduction, we think it is very important to examine whether change Tables 1 and 2, dysfunctional attitudes, negative (automatic) thoughts,
in two treatments is achieved through similar or differential pathways). rumination, worry and mindfulness skills were found to be associated
However, we think that conceptually it does not make sense to include with change in the majority of studies. Findings on the mediational
this as a first-order requirement for a mediator. In our view, the primary role of the other constructs that were investigated across studies are
more mixed. In general, approximately half of the studies examining a
1
If the abstract did not provide all the information necessary to assess in- and exclusion
4
criteria, the full article was consulted. Based on 24 studies; two studies did not report on this (Webb et al., 2013; Zettle et al.,
2
If a study used a mixed sample (e.g. depression and anxiety) but the main focus was 2011).
5
on depression and the majority of the sample was depressed, the study was included. Based on 8 studies; one study did not report on this (Smith et al., 2015)
3 6
Because we excluded studies as soon as they did not meet one of the inclusion criteria, Based on 33 studies; two studies did not report on this (Webb et al., 2013; Smith et al.,
the number of studies meeting multiple exclusion criteria is unknown. 2015).
 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107 99

Table 1
Characteristics and results of 35 identified studies aimed at identifying psychological mediators for (subclinical) depression, and the extent to which they meet requirements for process
research.

Study characteristics and results Requirements for process research

(NL)

(NL)

(NL)

(USA)

(USA)

(USA)

(USA)

(USA)

(NL)

(USA)

(continued on next page)

100 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107

Table 1 (continued)

(USA)

(USA)

(USA)

(USA)

(GER)

(USA)

(USA)

(AUS)

(USA)

(CHI)

(CAN)
 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107 101

Table 1 (continued)

(CAN)

(USA)

(USA)

(USA)

(UK)

(USA)

(USA)

(USA)

(NL)

(continued on next page)

102 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107

Table 1 (continued)

(UK)

(USA)

(USA)

(NL)

(USA)

ABBREVIATIONS: Column Headings: RCT = Design is Randomized-Controlled Trial (yes/no); n ≥ 40 = Sample size per treatment arm is at least 40 (yes/no); Control = Control Group (yes/no); Multiple
Mediators = Study included more than 1 potential mediator (yes/no); Temporality (ass n > 2) = Number of assessment in treatment phase (FU assessments are not included in count); Manipulation of
mediator = Manipulation of Mediator (yes/no); (+) = Present/Yes; (–) = Absent/No. Countries: AUS = Australia; CAN = Canada; CHI = China; GER = Germany; NL = the Netherlands; UK = United Kingdom;
USA = United States of America Interventions: ACT = Acceptance and Commitment Therapy; BA = Behavioural Activation; CBASP = Cognitive Behavioural Analysis System of Psychotherapy; CBSM =
Cognitive Behavioural Stress Management Treatment; CBT = Cognitive Behavioural Therapy; CM = clinical management; C-MBCT = Compassion focused MBCT; CT = Cognitive Therapy; DT = Dynamic
Therapy; ECS = Enhanced Community Standard; EFT-PE = Emotion-Focused Therapy, Process Experiential approach; IPT: Interpersonal Psychotherapy; MBCT = Mindfulness Based Cognitive Therapy;
MBSR = Mindfulness Based Stress Reduction; NST = Non-Directive Supportive Therapy; PHT = Pharmacotherapy; PST = Problem Solving Therapy; R-CBT = Rumination focused CBT; SBFT = Systematic
Behavioural Family Therapy; SE-CBT = Supportive-Expressive Cognitive Behavioural Therapy; TAU= Treatment-as-usual; WLC = Waiting-List Control. Outcome measures: BASIS = Behaviour and Symptom
Identification Scale; BDI = Beck Depression Inventory; BDI-II = Beck Depression Inventory II; BSI = Brief Symptom Inventory; BSQ = Before Session Questionnaire (items: symptom intensity and progress
towards goal); CDRS-R = Children’s Depression Rating Scale Revised; CES-D-(10) = Centre for Epidemiological Studies Depression Inventory (10 item version);; HRSD = Hamilton Rating Scale for
Depression; IDS-SR = Inventory of Depressive Symptoms, self-rating; K-SADS-P = Schedule for Affective Disorders and Schizophrenia for School-Age children, Present Episode Version; MFQ-C = Mood
and feelings Questionnaire Child; MHRSD = Modified Hamilton Rating Scale for Depression; PFS-IF = Psychodynamic Functioning Scale – Interpersonal Functioning Subscales; SCL-90-(R) = Symptoms
Checklist 90 (Revised). Statistical Method: SEM = structural equation modeling. Mediator variables: DAS = Dysfunctional Attitudes Scale. NOTES: a = Data come from RCT’s, but studies do not make use of
RCT design in analyses: Beevers et al. (2007), Shahar et al., 2004 and Watson et al. (2014) use a combined sample, and Ryba et al. (2014) only select patients allocated to BA. Therefore there is no control/
comparison group available in these studies; b = study was also included in the review by Garratt & Ingram (2007); c = Composite score of BDI, HRSD, Raskin Depression Scale; d = Used residualized
change scores; e = study sample consisted of patients diagnosed with depression (> 50%) and anxiety disorders; f = Description of the exact statistical procedure used in this study is lacking. Based on
reading the method section one could conclude that this is Baron & Kenny (1986) and Sobel test (1982); g = Composite score of BDI, SCL-90, HRDS, Global Assessment Scale (GAS) and Social Adjustment
Scale (SAS), as described by Blatt et al. (1996).

construct found evidence for mediation, whereas the other half did not as proposed by the MacArthur group (discussed in the Introduction)
find a relation between the mediator and outcome. When only focusing could have been a solution here.
on findings of C(BT) studies that examined treatment specific mediators
(n = 16), support for (partial) mediation was found in 63.3% of the cases. 3.2. A closer look at the value of these results
Exploration of the statistical methods of the 35 identified studies in-
dicated that early papers mainly examined the four basic steps of the As discussed by Kazdin (2007), after completing statistical media-
mediational model using linear regressions. The size of the indirect ef- tion analysis, one should return to the other criteria to assess the extent
fect was often examined with a Sobel (1982) test. However, as time to which they are met. The results of the assessment of requirements for
passed, a range of new (more sophisticated) statistical analyses tech- process research are presented for each individual study in the right
niques was observed. For example, mediational effects were now esti- panel of Table 1 and summarized in Table 3.
mated using multiple regression (ordinary least squares), logistic The majority of studies (74.3%) used an RCT design, and consequent-
regression, multilevel regression and structural equation modeling ly included one or more comparison groups. Interventions of interest
(SEM). The Sobel test was replaced with joint-significance testing were compared to a) other active treatments (psychological and/or
(MacKinnon et al., 2007) and bootstrapping (Preacher & Hayes, 2004, pharmacological; e.g. Blalock et al., 2008; DeRubeis et al., 1990;
2008). A closer look at the statistical methods showed that two studies Forman et al., 2012; Jacobs et al., 2009, Jacobs et al., 2014); b) treatment
(DeRubeis et al., 1990; Kolko, Brent, Baugher, Bridge, & Birmaher, 2000) as usual (e.g. Watkins et al., 2011); or c) non-active waiting-list control
could not finish their mediation analyses because the treatment condi- conditions (e.g. Fledderus, Bohlmeijer, Fox, Schreurs, & Spinhoven,
tions that were compared did not differ significantly with regard to out- 2013; Lo, Ng, Chan, Lam, & Lau, 2013; Shahar, Britton, Sbarra,
come. However, while DeRubeis et al. (1990) concluded that mediation Figueredo, & Bootzin, 2010; Smith et al., 2015). Since studies with an ac-
analysis was not possible because group differences were absent, Kolko tive control condition and those with a non-active (wait-list) compari-
et al. (2000) concluded that mediational effects of the proposed media- son group present different types of testing the significance of
tors were lacking. Re-analysing these data using the adapted guidelines mediators, we compared the results of studies with an active control
 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107 103

Table 2
Selection of significant mediators in the identified studies.

All studies (n = 35) Studies meeting 4 or more criteria (n = 17)

Examined Significant Examined Significant

– Dysfunctional attitudes 7 4 3 2
– Negative (automatic) thoughts 7 4 4 2
– Attributional style 3 1 2 1
– Behavioural concepts 6 3 4 3
– Mindfulness skills 4 3 3 2
– Rumination 5 4 2 1
– Worry 3 3 3 3
– Therapeutic alliance 3 1 1 1
Total 38 23 (61%) 22 15 (68%)

group (n = 16), with those including a non-active control group (n = mediated the effect of MBCT on depression severity, and that they did
11). Studies with a non-active control group showed relatively more so to the same extent.
statistically significant mediators as compared to studies with an active A closer look at the aspect of temporality identified three categories
contrast group (69.0 vs 41.3%). of studies. First of all, there were 12 studies that assessed mediator(s)
Four studies used data that originally came from RCTs, but did not and outcome more than twice during treatment, and were therefore
make use of the RCT design in their mediational analyses. They either able to make some kind of judgment about the temporal order of change
only selected patients allocated to one particular condition (Ryba, (e.g. DeRubeis et al., 1990; Fledderus et al., 2013; Jacobs et al., 2009;
Lejuez, & Hopko, 2014), or merged the various intervention groups into Kwon & Oei, 2003; Warmerdam et al., 2010). Two of these twelve stud-
one combined sample (Beevers, Wells, & Miller, 2007; Shahar, Blatt, ies even assessed mediators and outcome on a session-by-session basis
Zuroff, Krupnick, & Sotsky, 2004; Watson, Steckley, & McMullen, 2014). (Forman et al., 2012; Ryba et al., 2014). The second group consisted of
As a result, there was no control/comparison group available. In addition, studies that only included pre- and post-treatment assessments. By
as can be seen in Table 3, the number of studies including a control group assessing processes and outcomes only at pre- and post-treatment one
is higher than the number of studies with an RCT design. This can be can say that change in a mediator indeed correlates with, explains a cer-
explained by the fact that one study compared two treatments in a tain amount of variance, or predicts change in outcome, but not whether
non-randomized design (Klug, Henrich, Filipiak, & Huber, 2012). one process precedes the other. For example, Quilty, McBride, and
Table 3 furthermore shows that two thirds of the selected studies in- Bagby (2008) found in their study that a decrease in dysfunctional atti-
cluded sample sizes of N 40 participants per condition. This was even the tudes was associated with a decrease in depression severity in CBT. This
case in several RCTs with three or four arms (e.g. Jacobs et al., 2009; is in line with cognitive theory of depression. However, no conclusions
Jacobs et al., 2014; Stice, Rohde, Seeley, & Gau, 2010). However, the about temporality could be drawn because mediators and outcome
small sample sizes in various other studies show that power can still measures were only measured twice at the same assessments. Similarly,
be an issue in this type of research, also in relatively recent studies the fact that Allart-van Dam et al. (2003) found that changes in depres-
(e.g. Ryba et al., 2014; Watkins et al., 2011; Zettle, Rains, & Hayes, sive cognition and self-esteem were significant mediators of depressive
2011). The smallest sample was found in a study by Gaynor and symptoms following a coping with depression course, is of less value be-
Harris (2008), who conducted single participant assessment of media- cause they only used two assessment points. Other examples can be
tors in four depressed adolescents. One study explicitly compensated found in Table 1. A third category consisted of studies that did in-
for the small sample size and low power by conducting mediation anal- clude more than two assessment points, but not within the active
yses with the therapy groups combined (Watson et al., 2014). phase of treatment. For example, Kuyken et al. (2010), included a
Almost 80% of studies included more than one mediator in their de- total of three assessment points, but one of them was at 15 month
sign. Some studies included several separate potential processes of follow-up, leaving only 2 assessments during treatment (baseline
change (e.g. Allart-van Dam, Hosman, & Hoogduin, 2003; Kaufman, and post-treatment). A similar approach was used by Toth et al.
Rohde, Seeley, Clarke, & Stice, 2005; Warmerdam et al., 2010), whereas (2013). Even though this is very informative regarding to the knowl-
others examined subscales of the same construct (Blalock et al., 2008; edge on mediators of sustained treatment effects, it will not help to
Lewis et al., 2009). However, even when multiple mediators were in- reveal mechanisms during treatment. Lastly, none of the identified
cluded in a study, they were often analysed individually. Only a small studies used an approach in which the proposed mediator was ex-
number of studies looked at the relative importance and collaboration perimentally manipulated.
between several potential mechanisms. For example, the study by
Batink, Peeters, Geschwind, van Os, & Wichers (2013) indicated that 3.2.1. The criteria in concert
even though both positive affect, as well as negative affect played a sub- Since satisfaction of each criterion increases the strength of the argu-
stantial mediating role in the reduction of depressive symptoms during ment for the operation of a mediator, further interpretation of findings
MBCT, the effect of the first was larger compared to the latter. Shahar et should be based on concerted action between these criteria. We there-
al. (2010) also included several potential mediators in one model and fore also looked at the total number of criteria met by each study. An
showed that changes in mindfulness and changes in brooding both overview is given in Fig. 1. As can be seen in the figure, not one study
meets all criteria.
Table 3 Four studies scored 5 out of 6 and seem to be the most promising
Number (%) of studies meeting requirements for process research (n = 35). with regard to meeting the various criteria. Forman et al. (2012) exam-
Requirement n studies (%) ined the mediating role of theorized mechanisms in ACT and CT (utiliza-
tion of cognitive acceptance vs. change, utilization of affective
– RCT, yes, n (%) 26 (74.3)
– Control group, yes, n (%) 27 (77.1) acceptance vs. change, dysfunctional thinking, cognitive defusion and
– Sample size per condition ≥40, yes, n (%) 23 (65.7) committed action). They found that treatment group moderated the
– Multiple mediators, yes, n (%) 27 (77.1) mediating effects of both cognitive and affective changes. More specifi-
– Assessment of temporality, yes, n (%) 12 (34.3) cally, cognitive techniques facilitated outcome for those receiving CT,
– Manipulation of mediator/experiment, yes, n (%) 0 (0.0)
whereas utilization of psychological acceptance strategies facilitated
104 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107

In addition, 13 studies met 4 out of 6 criteria. As can be seen in

Table 1, the combination of criteria that were met was different for
the various studies. Apart from the manipulation criterion, many studies
did not meet the requirement of temporality. The remaining 18 studies
met b4 criteria. Remarkably, the two studies meeting only one criterion
(Kwon & Oei, 2003; Ryba et al., 2014), met the temporality criterion,
which was lacking in many of the other studies.
The question that remains is what is left of the evidence when only
taking ‘high quality’ studies into consideration (i.e. studies that meet
≥4 criteria). Results of studies meeting 4 or more criteria (n = 17) are
displayed in the right panel of Table 2. As can be seen in the table, the
relative percentage of studies finding significant mediators is slightly
higher than when all studies are taken into account (68% vs 61%). How-
ever, results should be interpreted with caution given the relatively
Fig. 1. The criteria in concert: number of studies per number of criteria met.
small number of studies per potential mechanism.

outcome in ACT. Results of this study are promising since they are ob- 4. Discussion
tained in a large RCT (n = 174) with repeated assessments (before
each session) of multiple mediators and outcomes. It has to be noted We provided a systematic empirical update and critical evaluation of
however, that they included a mixed sample of patient with anxiety the current status of research aimed at identifying a variety of psycho-
and depression, and did not control for the influence diagnosis. Stice logical mediators in various forms of psychotherapy for depression.
et al. (2010) randomized 341 teens with elevated levels of depression With this we wanted to learn more about the magnitude and relevance
to group CBT, Group Supportive Expressive therapy (SET), Cognitive Be- of the existing body of research and map out necessities for future stud-
havioural Bibliotherapy or assessment-only control, and examined the ies. We summarized study characteristics and results of 35 relevant em-
mediating role of theorized processes of change of CBT (negative cogni- pirical studies that were identified in a systematic literature search, and
tions/pleasant activities) and SET (emotional expression/loneliness). discussed the extent to which these studies meet several important re-
Separate analyses were conducted for each of the active treatments, in quirements for mechanism research. The selected studies examined a
which each treatment was contrasted to the non-active control. The re- total of 39 potential mediators in 12 different treatment modalities.
sults on the Group CBT intervention indicated the presence of a media- Conclusions about the mediational role of the various constructs that
tor: the treatment reduced depressive symptoms, negative cognitions, were examined across studies were mixed, potentially due to a large
and increased pleasant activities. Furthermore, change in these process- variation in research questions, methodology and quality of studies.
es predicted change in depression, and intervention effects became However, despite this variation, several processes (e.g. dysfunctional at-
weaker when controlling for change in the processes. However, after titudes, negative (automatic) thoughts, rumination, worry and mindful-
examination of the sequence of changes, it was found that change in de- ness skills) were associated with change in the majority of studies
pression occurred before change in the mediator. Therefore it was con- reviewed, and therefore warrant further examination. In doing this, it
cluded that changes in theorized processes did not mediate the would be important to also take the specificity-hypothesis into account.
intervention effects. This illustrates the importance of including the as- Not so much as a requirement for mediators, but in order to broaden our
pect of temporality. A similar (but less strong) pattern was found for overall knowledge about the processes associated with therapeutic
SET. Quilty, Dozois, Lobo, and Bagby (2014) examined the temporal dy- change. Studies with a non-active control group showed relatively
namics and causal role of cognitive structure and processing in CBT more significant mediators than studies with an active control group.
(n = 54) vs. pharmacotherapy (n = 50) for depression. The authors in- None of the identified studies met all requirements for tests of treat-
cluded multiple mediator measures and outcomes that were assessed ment mediation, mainly because studies were unable to assess the tem-
at various points before, during and after treatment. Data were poral relationship between change in the mediator and change in
analysed using modern statistical methods. In spite of a well-considered outcome, and because none of the studies used an approach in which
design, the evidence for the mediational role of the investigated the proposed mediator was experimentally manipulated. Of course,
constructs was weak. Only two out of 14 subscales exhibited (partial) one can question the prominence of this latter criterion, as the external
mediation on one of the outcome measures. Effects did not seem to validity of experiments that manipulate a proposed mechanism in isola-
be specific for CBT. Warmerdam et al. (2010) studied the mediating tion, keeping everything else constant, might be limited.
role of dysfunctional attitudes, worry, negative problem orientation, When comparing our findings to those of previous reviews in the
and feelings of control in online CBT and PST for depression. A total field (e.g. Garratt et al, 2007; Johansson & Høglend, 2007), it can be con-
of 263 participants were randomly allocated to one of the two cluded that some advances have been made in theoretical consensus
active treatment conditions, or to a waiting-list condition. Measures about necessities for this type of research, and in the degree of sophisti-
were taken at three points over the course of treatment. Similarly cation that researchers bring to research on mediators. More and more
to Stice et al. (2010) active conditions were contrasted to the WLC attention is paid to the aspect of temporality, sample size, and the inclu-
condition. Warmerdam and colleagues found support for the sion of multiple processes in one study. Nevertheless, the empirical state
notion that the mechanisms of interest played a mediating role of affairs has only shown little progress in the past decade. Research is still
in both CBT as well as PST. Multiple mediation analysis showed that – heterogeneous and often unsatisfactory in methodological regard.
in both groups – reduction in depression was mostly explained Probably the biggest challenge in research aimed at identifying mediators
by improvement in worrying, perceived control and a negative is demonstrating the causal relation between change in the mediator and
problem orientation. However, since most of the total improvement change in depression severity. As a result, after more than three decades
had already taken place before the mid-treatment assessment of process research focused on depression treatment, there is still no
(5 months) – leaving only little room for later change – the authors clear-cut empirical explanation for psychotherapeutic change.
were not able to differentiate between cause and effect. So in spite of Demonstrating causality is difficult though, even in studies that are
a suitable repeated measures design, and promising results, they were designed to explain therapeutic change in terms of causal processes.
not able to discern the temporal relation necessary to identify a mecha- First of all, determining the best timing and spacing of observations to
nism of change. capture the critical point of change is a difficult and delicate matter,
 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107 105

especially when there is no prior information available about the speed influence of the choice of the contrast group. Moreover, apart from tra-
and shape of change. One needs to balance the most optimal study de- ditional designs to examine processes of change, alternative designs -
sign, with the burden for patients, and the risk of measurement artefacts including e.g. experimental manipulations, component analyses, and
when making too many demands for data (Longwell & Truax, 2005). Sequential Multiple Assignment Randomized Trials (SMART) designs -
Furthermore, research designs are often based on the assumption that should be considered as well. After the identification of processes that
change is gradual and linear. However, various studies have shown are a linking pin between treatment and outcome, further analyses
that change often happens sudden, rather than gradually over the should examine whether these processes play a role in only one treat-
course of treatment (see review of Aderka, Nickerson, Bøe, & ment, or are relevant for other treatments as well. In doing all of this,
Hofmann, 2011 for more details). If therapeutic change indeed occurs it is important that researchers invest in the development of a uniform
suddenly (e.g. the ‘aha-experience’) it might be very difficult to capture research language, and standardized assessment- and research proto-
this moment, let alone to assess the temporal relation between change cols. This will make it easier to compare results across studies, and inte-
in the mechanism and change in symptoms. grate findings into broader knowledge.
Research aimed at identifying the active ingredients of psychothera- Furthermore, identifying and understanding mediators relies on
py for depression would benefit from a further refinement of research theory about mechanisms of change. Statistical tests of mediation
methods to disentangle mechanisms of change. Table 4 gives an over- are tools, silent as to content, and without theory we cannot answer
view of several recommendations for future research. In short, future the questions that we are still confronted with despite multiple de-
studies should focus on establishing a more fine-grained analysis of cades of research. Without theory, we do not know which mecha-
the exact shape of change. Studies should include multiple measures nisms might play a role and should be tested. Theories on
of potential (specific and non-specific) process measures and outcomes mechanisms of change do exist, but often do not specifically account
in well-planned temporal research designs paying special attention to for the interplay between multiple (specific and non-specific) pro-
the timing of assessments and within-patient variances. This is not cesses. Furthermore, little progress has been made during the last
only relevant in the light of examining the causal relation between decades on the theoretical level: basically, we are still testing the
change in the mediator and change in outcome, but could also provide same mechanisms that were proposed 20 years ago. It is rather dis-
more insight in the differential patterns of change of two treatments appointing that almost two decades after these points were raised
that overall have comparable effects. Experience Sampling Methods (e.g. Kazdin, 1999; Kurtines & Silverman, 1999) we see very little
(ESM) might be promising in this regard. With regard to the choice of progress in the field. Therefore, apart from advances in research
mediator variables it is important to examine the role of both theorized methods, the field urgently needs further development of theories
(specific and non-specific) processes, as well as of processes that theo- of therapeutic change. When constructing and evaluating theoretical
retically might not mediate the relation between treatment and out- models of change, it would be useful to not only look at the theoret-
come since this can serve as an important tool to further examine ical mediators of a particular treatment, but also to consider how
whether treatments work for the hypothesized reasons, or due to other treatments would be expected to affect these mediators and
other processes. Furthermore, researchers should invest in the develop- how the mediators would be expected to affect the outcome.
ment and evaluation of mediator measures. In particular, fundamental However, even with well-considered theoretical frameworks and
research on the validity of process measures should progress. In addi- optimal research designs, explaining psychotherapeutic change re-
tion, it is important that researchers use sophisticated statistical mains a challenge. Psychotherapy for depression is a complex,
methods for the analysis of change and pay attention to the potential multi-dimensional phenomenon that might work through interplay
of multiple mechanisms on several levels (physiological, affective,
behavioural and cognitive aspects). Psychotherapeutic change
Table 4 might therefore consist of a complicated chain of events on these dif-
Recommendations for future research aimed at identifying mediators.
ferent levels. In addition, it is possible that active components of
Potential mediators therapy and their associated mechanisms of change work differently
– Invest in further development of theories of therapeutic change. at different points in time and differ between (subgroups of) de-
– Use theory to select multiple specific and non-specific potential mediators.
pressed patients. With this in mind, psychotherapeutic change
– Include processes that would falsify the theory as well.
– Provide a clear description of each process that is included. might even be too complex to be explained in relatively simple caus-
– Use mediator measures that have shown to be psychometrically valid. al models of psychological change. If this is the case, psychological
– Invest in the evaluation and further development of (implicit) mediator research designs might never be able to explain all aspects of thera-
measures. peutic change. However, it would make it a lot easier to understand
– Use multiple sources of information (self-report, clinician rated, independent
rater, and behavioural and biological measurements).
why research so far has not led to clear-cut empirical explanations
of how psychotherapy for depression works.
Study design
– (Multi-site) RCTs with a control group, preferably also including a non-active
arm.
Appendix A. Key-term scheme for database search
– Include a fine grained temporal design, especially in the early phase of
treatment. ‘Psychotherapy’, ‘Psychotherapies’, ‘Psychological Treatment(s)/
– Justify the timing and spacing of observations Intervention(s)’, ‘Interpersonal (Psycho)therapy’, ‘(Mindfulness-
– Invest in development of alternative research designs including experimental
based) Cognitive (Behavio(u)ral) Therapy’, ‘Psychodynamic/analytic
manipulations and component analyses.
Therapy’, ‘Client-Centered Therapy’, ‘Behavio(u)ral Activation’,
Analyses ‘Acceptance Commitment Therapy’; ‘Mechanisms of Change/Action’,
– Use modern statistical analysis methods to examine change over time and
‘Working Mechanisms (of psychotherapy)’, ‘Processes of therapy’,
mediation.
– Focus on statistical significance but also on the clinical meaning of changes. ‘Process Research’, ‘Change’; ‘Mediation’, ‘Mediator’, ‘Mediating effects’;
– Examine the unique influence of each mediator, as well as their interactions. ‘Depression’, ‘Major Depressive Disorder’, ‘Dysthymia’, ‘Dysthymic
– Perform analysis on group level; but also examine subgroups, and individual Disorder’
trajectories.

Reporting Appendix B. In & exclusion criteria

– Invest in 1 research language and standard guidelines for reporting
mechanisms. • Published in English in Peer-reviewed Journal
– Replicate studies and publish negative data as well.
• Empirical Research report (no review/theoretical paper/commentary)
106 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107

• Focus = Psychological mechanisms of treatment for Depression Hill, A. B. (1965). The environment and disease: Association or causation? Proceedings of
• Population = Diagnosis/Symptoms of Depression the Royal Society of Medicine, 58, 295–300.
• Intervention = (Evidence-based) psychotherapy Hollon, S. D., & DeRubeis, R. J. (2009). Mediating the effects of cognitive therapy for de-
• Including clinical outcome measure for depression pression. Cognitive Behaviour Therapy, 38(S1), 43–47.
• Performing Statistical Mediation analysis Hoyle, R. H., & Kenny, D. A. (1999). Statistical power and tests of mediation. In R. H. Hoyle
(Ed.), Statistical strategies for small sample research. Newbury Park: Sage.
Jacobs, R. H., Silva, S. G., Reinecke, M. A., Curry, J. F., Ginsburg, G. S., Kratochvil, C. J., &
March, J. S. (2009). Dysfunctional attitudes scale perfectionism: A predictor and par-
References tial mediator of acute treatment outcome among clinically depressed adolescents.
Journal of Clinical Child and Adolescent Psychology, 38(6), 803–813.
Aalderen, v. J. R., Donders, A. R. T., Giommi, F., Spinhoven, P., Barendregt, H. P., & Speckens, Jacobs, R. H., Becker, S. J., Curry, J. F., Silva, S. G., Ginsburg, G. S., Henry, D. B., & Reinecke, M.
A. E. M. (2012). The efficacy of mindfulness-based cognitive therapy in recurrent de- A. (2014). Increasing positive outlook partially mediates the effect of empirically sup-
pressed patients with and without a current depressive episode: A randomized con- ported treatments on depression symptoms among adolescents. Journal of Cognitive
trolled trial. Psychological Medicine, 42(5), 989–1001. Psychotherapy, 28(1), 3–19.
Aderka, I. M., Nickerson, A., Bøe, H. J., & Hofmann, S. G. (2011). Sudden gains during psy- Johansson, P., & Høglend, P. (2007). Identifying mechanisms of change in psychother-
chological treatments of anxiety and depression: A meta-analysis. Journal of apy: Mediators of treatment outcome. Clinical Psychology & Psychotherapy, 14(1),
Consulting and Clinical Psychology, 80(1), 93–101. 1–9.
Allart-van Dam, E., Hosman, C. M. H., & Hoogduin, C. A. L. (2003). The coping with depres- Johansson, P., Høglend, P., Ulberg, R., Amlo, S., Marble, A., Bøgwald, K. -P., ... Heyerdahl, O.
sion course: Short-term outcomes and mediating effects of a randomized controlled (2010). The mediating role of insight for long-term improvements in psychodynamic
trial in the treatment of subclinical depression. Behavior Therapy, 34(3), 381–396. therapy. Journal of Consulting and Clinical Psychology, 78(3), 438–448.
Arbuckle, J. L. (1999). AMOS: A structural equation modeling software. Chicago, IL: Kaufman, N. K., Rohde, P., Seeley, J. R., Clarke, G. N., & Stice, E. (2005). Potential
SmallWaters Corporation. mediators of cognitive-behavioral therapy for adolescents with comorbid
Arbuckle, J. L. (2005). Amos 6.0 user's guide. Spring House, PA: Amos Development major depression and conduct disorder. Journal of Consulting and Clinical
Corporation. Psychology, 73(1), 38–46.
Backenstrass, M., Schwarz, T., Fiedler, P., Joest, K., Reck, C., Mundt, C., & Kronmueller, K. -T. Kazdin, A. E. (1999). Current (lack of) status of theory in child and adolescent psychother-
(2006). Negative mood regulation expectancies, self-efficacy beliefs, and locus of con- apy research. Journal of Clinical Child Psychology, 28(4), 533–543.
trol orientation: Moderators or mediators of change in the treatment of depression? Kazdin, A. E. (2007). Mediators and mechanisms of change in psychotherapy research.
Psychotherapy Research, 16(2), 250–258. Annual Review of Clinical Psychology, 3, 1–27.
Baron, R. M., & Kenny, D. A. (1986). The moderator - mediator variable distinction in so- Kazdin, A. E. (2009). Understanding how and why psychotherapy leads to change.
cial psychological research: Conceptual, strategic, and statistical considerations. Psychotherapy Research, 19(4–5), 418–428.
Journal of Personality and Social Psychology, 51(6), 1173–1182. Kazdin, A. E., & Nock, M. K. (2003). Delineating mechanisms of change in child and ado-
Batink, T., Peeters, F., Geschwind, N., van Os, J., & Wichers, M. (2013). How does MBCT for lescent therapy: Methodological issues and research recommendations. Journal of
depression work? Studying cognitive and affective mediation pathways. PloS One, Child Psychology and Psychiatry, 44(8), 1116–1129.
8(8), 1–13. Kenny, D. A. (1979). Correlation and causality. New York, NY: Wiley.
Beck, A. T., Ward, C. H., Meldelson, M., Mock, J., & Erbauch, J. (1961). An inventory for Klug, G., Henrich, G., Filipiak, B., & Huber, D. (2012). Trajectories and mediators of change
measuring depression. Archives of General Psychiatry, 4, 561–571. in psychoanalytic, psychodynamic, and cognitive behavioral therapy. Journal of the
Beck, A. T., Steer, R., & Brown, G. K. (1996). Beck depression inventory II: Manual. Boston, American Psychoanalytic Association, 60(3), 598–605.
MA: Hartcourt Brace. Kolko, D. J., Brent, D. A., Baugher, M., Bridge, J., & Birmaher, B. (2000). Cognitive and family
Beevers, C. G., Wells, T. T., & Miller, I. W. (2007). Predicting response to depression treat- therapies for adolescent depression: Treatment specificity, mediation, and modera-
ment: The role of negative cognition. Journal of Consulting and Clinical Psychology, tion. Journal of Consulting and Clinical Psychology, 68(4), 603–614.
75(3), 422–431. Kraemer, H. C., Stice, E., Kazdin, A. E., Offord, D., & Kupfer, D. J. (2001). How do risk factors
Blalock, J. A., Fouladi, R. T., Cinciripini, P. M., Markowitz, J. C., Klein, D. N., Rothbaum, B. O., work together? Mediators, moderators, and independent, overlapping and proxy risk
... McCullough, J. P. (2008). Cognitive and behavioral mediators of combined pharma- factors. American Journal of Psychiatry, 158(6), 848–883.
cotherapy and psychotherapy of chronic depression. Cognitive Therapy and Research, Kraemer, H. C., Wilson, G. T., Fairburn, C. G., & Agras, W. S. (2002). Mediators and moder-
32(2), 197–211. ators of treatment effects in randomized clinical trials. Archives of General Psychiatry,
Blatt, S. J., Zuroff, D. C., Quinlan, D. M., & Pilkonis, P. (1996). Interpersonal factors in 59, 877–883.
brieftreatment of depression: Further analyses of the National Institute of Mental Kurtines, W. M., & Silverman, W. K. (1999). Emerging views of the role of theory. Journal
HealthTreatment of Depression Collaborative Research Program. Journal of of Clinical Child Psychology, 28(4), 558–562.
Consulting and Clinical Psychology, 64, 162–171. Kuyken, W., Watkins, E., Holden, E., White, K., Taylor, R. S., Byford, S., ... Dalgleish, T.
Collins, L. M., & Graham, J. W. (2002). The effect of the timing and temporal spacing of ob- (2010). How does mindfulness-based cognitive therapy work? Behaviour Research
servations in longitudinal studies of tobacco and other drug use: Temporal design and Therapy, 48, 1105–1112.
considerations. Drug and Alcohol Dependence, 68, 85–96. Kwon, S., & Oei, T. P. S. (2003). Cognitive change processes in a group cognitive behavior
Cuijpers, P., & van Straten, A. (2011). New psychotherapies for mood and anxiety disor- therapy of depression. Journal of Behavior Therapy and Experimental Psychiatry, 34,
ders: Necessary innovation or waste of resources? Canadian Journal of Psychiatry, 73–85.
56(4), 251–252. Laurenceau, J. -P., Hayes, A. M., & Feldman, G. C. (2007). Some methodological and statis-
Cuijpers, P., van Straten, A., Andersson, G., & van Oppen, P. (2008). Psychotherapy for de- tical issues in the study of change processes in psychotherapy. Clinical Psychology
pression in adults: A meta-analysis of comparative outcome studies. Journal of Review, 27(6), 682–695.
Consulting and Clinical Psychology, 76(6), 909–922. Lewis, C. C., Simons, A. D., Silva, S. G., Rohde, P., Small, D. M., Murakami, J. L., ... March, J. S.
DeRubeis, R. J., Evans, M. D., Hollon, S. D., Garvey, M. J., Grove, W. M., & Tuason, V. B. (2009). The role of readiness to change in response to treatment of adolescent de-
(1990). How does cognitive therapy work? Cognitive change and symptom change pression. Journal of Consulting and Clinical Psychology, 77(3), 422–428.
in cognitive therapy and pharmacotherapy for depression. Journal of Consulting and Lo, H. H. M., Ng, S. M., Chan, C. L. W., Lam, K. F., & Lau, B. H. P. (2013). The Chinese med-
Clinical Psychology, 58(6), 862–869. icine construct “stagnation” in mind–body connection mediates the effects of mind-
Fledderus, M., Bohlmeijer, E. T., Fox, J., Schreurs, K. M. G., & Spinhoven, P. (2013). The role fulness training on depression and anxiety. Complementary Therapies in Medicine,
of psychological flexibility in a self-help acceptance and commitment therapy inter- 21(4), 348–357.
vention for psychological distress in a randomized controlled trial. Behaviour Longmore, R. J., & Worrell, M. (2007). Do we need to challenge thoughts in cognitive be-
Research and Therapy, 51, 142–151. havior therapy? Clinical Psychology Review, 27(2), 173–187.
Forman, E. M., Chapman, J. E., Herbert, J. D., Goetter, E. M., Yuen, E. K., & Moitra, E. (2012). Longwell, B. T., & Truax, P. (2005). The differential effects of weekly, monthly, and bi-
Using session-by-session measurement to compare mechanisms of action for acceptance monthly administrations of the Beck depression inventory-II: Psychometric proper-
and commitment therapy and cognitive therapy. Behavior Therapy, 43(2), 341–354. ties and clinical implications. Behaviour Therapy, 36, 265–275.
Garratt, G., Ingram, R. E., Rand, K. L., & Sawalani, G. (2007). Cognitive processes in cogni- MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. New York, NY: Law-
tive therapy: Evaluation of the mechanisms of change in the treatment of depression. rence Erlbaum Associates.
Clinical Psychology, 14, 224–239. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A com-
Gaynor, S. T., & Harris, A. (2008). Single-participant assessment of treatment mediators: parison of methods to test mediation and other intervening variable effects.
Strategy description and examples from a behavioral activation intervention for de- Psychological Methods, 7, 83–104.
pressed adolescents. Behavior Modification, 32(3), 372–402. MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence limits for the indirect
Haaga, D. A. F., & Stiles, W. B. (2000). Randomized clinical trials in psychotherapy re- effect: Distribution of the product and resampling methods. Multivariate Behavioral
search: Methodology, design and evaluation. In C. R. Snyder, & R. E. Ingram (Eds.), Research, 39(1), 99–128.
Handbook of psychological change: Psychotherapy processes and practices for the 21st MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual Review of
century (pp. 14–39). New York, NY: Wiley. Clinical Psychology, 58, 593–614.
Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery and Murphy, R., Cooper, Z., Hollon, S. D., & Fairburn, C. G. (2009). How do psychological treat-
Psychiatry, 23, 56–62. ments work? Investigating mediators of change. Behaviour Research and Therapy,
Haubert, L. C., & Dobson, K. S. (2007). Treatment of depression and mechanisms of 47(1), 1–5.
change: Strengthening the links among theory, research and practice. Clinical Muthén, L. K., & Muthén, B. O. (2001). Mplus user's guide (2 ed.). Los Angeles, CA: Muthén
Psychology: Science and Practice, 14, 247–251. & Muthén.
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: Muthén, L. K., & Muthén, B. O. (2007). Mplus user's guide (6 ed.). Los Angeles, CA: Muthén
A regression-based approach. New York: Guilford Press. & Muthén.
 L.H.J.M. Lemmens et al. / Clinical Psychology Review 50 (2016) 95–107 107

Preacher, K., & Hayes, A. (2004). SPSS and SAS procedures for estimating indirect effects in Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equa-
simple mediation models. Behavior Research Methods, Instruments, & Computers, tion models. In S. Leinhardt (Ed.), Sociological methodology (pp. 290–312). Washing-
36(4), 717–731. ton DC: American Sociological Association.
Preacher, K., & Hayes, A. (2008). Asymptotic and resampling strategies for assessing and Stice, E., Rohde, P., Seeley, J. R., & Gau, J. M. (2010). Testing mediators of intervention ef-
comparing indirect effects in multiple mediator models. Behavior Research Methods, fects in randomized controlled trials: An evaluation of three depression prevention
40(3), 879–891. programs. Journal of Consulting and Clinical Psychology, 78(2), 273–280.
Quilty, L. C., McBride, C., & Bagby, R. M. (2008). Evidence for the cognitive mediational Toth, S. L., Rogosch, F. A., Oshri, A., Gravener-Davis, J., Sturm, R., & Morgan-López, A. A.
model of cognitive behavioural therapy for depression. Psychological Medicine, 38, (2013). The efficacy of interpersonal psychotherapy for depression among economi-
1531–1541. cally disadvantaged mothers. Development and Psychopathology, 25, 1065–1078.
Quilty, L. C., Dozois, D. J. A., Lobo, D. S. S., & Bagby, R. M. (2014). Cognitive structure and Vittengl, J. R., Clark, L. A., Thase, M. E., & Jarrett, R. B. (2014). Are improvements in cogni-
processing during cognitive behavioral therapy vs. pharmacotherapy for depression. tive content and depressive symptoms correlates of mediators during acute-phase
International Journal of Cognitive Therapy, 7(3), 235–250. cognitive therapy for recurrent major depressive disorder? International Journal of
Ryba, M. M., Lejuez, C. W., & Hopko, D. R. (2014). Behavioral activation for depressed Cognitive Therapy, 7(3), 255–271.
breast cancer patients: The impact of therapeutic compliance and quantity of activi- Wampold, B. E., Mondin, G. W., Moody, M., Stich, F., Benson, K., & Ahn, H. (1997). A meta-
ties completed on symptom reduction. Journal of Consulting and Clinical Psychology, analysis of outcome studies comparing bona fide psychotherapies: empirically, “All
82(2), 325–335. Must Have Prizes”. Psychological Bulletin, 122(3), 203–215.
Sasso, K. E., Strunk, D. R., Braun, J. D., DeRubeis, R. J., & Brotman, M. A. (2015). Identifying Warmerdam, L., van Straten, A., Jongsma, J., Twisk, J., & Cuijpers, P. (2010). Online cogni-
moderators of the adherence-outcome relation in cognitive therapy for depression. tive behavioral therapy and problem-solving therapy for depressive symptoms: Ex-
Journal of Consulting and Clinical Psychology, 83(5), 976–984. ploring mechanisms of change. Journal of Behavior Therapy and Experimental
Schlesselman, J. J. (1982). Case-control studies: Design, conduct and analysis. New York, NY: Psychiatry, 41, 64–70.
Oxford University Press. Watkins, E. D., Mullan, E., Wingrove, J., Rimes, K., Steiner, H., Bathurst, N., ... Scott, J.
Shahar, G., Blatt, S. J., Zuroff, D. C., Krupnick, J. L., & Sotsky, S. M. (2004). Perfectionism im- (2011). Rumination-focused cognitive-behavioural therapy for residual depression:
pedes social relations and response to brief treatment for depression. Journal of Social Phase II randomised controlled trial. The British Journal of Psychiatry, 199, 317–322.
and Clinical Psychology, 23(2), 140–154. Watson, J. C., Steckley, P. L., & McMullen, E. J. (2014). The role of empathy in promoting
Shahar, B., Britton, W. B., Sbarra, D. A., Figueredo, A. J., & Bootzin, R. R. (2010). Mechanisms change. Psychotherapy Research, 24(3), 286–298.
of change in mindfulness-based cognitive therapy for depression: Preliminary evi- Webb, C. A., Kertz, S. J., Bigda-Peyton, J. S., & Björgvinsson, T. (2013). The role of pretreat-
dence from a randomized controlled trial. International Journal of Cognitive Therapy, ment outcome expectancies and cognitive–behavioral skills in symptom improve-
3(4), 402–418. ment in an acute psychiatric setting. Journal of Affective Disorders, 149(1–3), 375–382.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental stud- Zettle, R. D., Rains, J. C., & Hayes, S. C. (2011). Processes of change in acceptance and com-
ies: New procedures and recommendations. Psychological Methods, 7(4), 422–445. mitment therapy and cognitive therapy for depression: A mediation reanalysis of
Smith, P., Scott, R., Eshkevari, E., Jatta, F., Leigh, E., Harris, V., ... Yule, W. (2015). Zettle and Rains. Behavior Modification, 35(3), 265–283.
Computerised CBT for depressed adolescents: Randomised controlled trial.
Behaviour Research and Therapy, 73, 104–110.