ChapterPDF Available

Barber, J. P., Muran, J.C., McCarthy, K.S., Keefe, R.J. (2013). Research on Psychodynamic Therapies. In M. J. Lambert (Ed.). Bergin and Garfield's Handbook of Psychotherapy and Behavior Change (6th ed.) (pp. 443-494). New-York, NY: John Wiley & Sons, Inc.


Abstract and Figures

This chapter summarizes the empirical research on psychodynamic therapies focusing on 3 main questions: 1) what is the evidence for the efficacy of dynamic therapy? 2) what do we know about the role of the therapeutic alliance in dynamic therapy? 3) what do we know about the main mechanisms of change in dynamic therapy?
Content may be subject to copyright.
Barber, J. P., Muran, J.C., McCarthy, K.S., Keefe, R.J. (2013). Research on Psychodynamic
Therapies. In M. J. Lambert (Ed.). Bergin and Garfield's Handbook of Psychotherapy and Behavior
Change (6th ed.) (pp. 443-494). New-York, NY: John Wiley & Sons, Inc.
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 443
Once psychoanalysis or psychodynamic therapy ruled the earth. Unlike the dinosaurs, it did not
disappear but rather sprouted many variations and new offspring. Today those offspring have
forgotten everything about their origins.
A chapter on dynamic therapy (DT) has been
omitted from some editions of this handbook.
Not including one may have been seen as some-
what consistent with the decrease in the place and
importance of psychodynamic psychotherapies
in academic psychology, especially in English-
speaking countries. However, we think the under-
representation of DT is unfortunate, as its main
ideas are at the foundation of many forms of
psychotherapy, including recently developed
ones (e.g., Summers & Barber, 2009). Ignoring
and neglecting the historical legacy and clinical
wisdom of dynamic therapy may adversely impact
clinical care (e.g., the role of emotions, implicit
cognitions, and defenses). It is quite clear (e.g.,
Summers & Barber, 2009) that many forms of
therapy evolved from psychoanalysis, including
cognitive and humanistic therapies. Some of the
important theorists behind these relatively newer
approaches were trained as psychoanalysts (e.g.,
A. T. Beck, Fritz Perls, Albert Ellis) and for vari-
ous reasons were dissatisfied with major aspectsof
psychoanalysis. Thus, these influential individuals
went on to make improvements over psychoanal-
ysis and developed their own version of therapy.
In their attempts to define their new approach to
treatment, they emphasized what was different
from psychoanalysis and perhaps minimized what
they took with them from those therapies (e.g.,
the importance of the therapeutic relationship and
encouraging a more realistic view of the world).
However, more relevant to the present volume
is the awakening of psychodynamically oriented
researchers in the past two decades. Specifically,
there has been an increase in empirical evidence
for the efficacy of these DT for specific disorders
as well as investigations on how and for whom
DT might be helpful, as we will show.
Characteristics of DT
This chapter covers approaches to psychotherapy
generally referred to as psychodynamic or ana-
lytic. DT—as well as cognitive-behavior therapy
(CBT) for that matter represents a family of
therapies. As in all large families, the degrees of
divergence and agreement vary quite a bit. Some
members of the families are not even on speaking
terms and sometimes even speak different lan-
guages. Thus, by necessity, this chapter presents
a narrow representation of the psychodynamic
family of therapies. Under this large umbrella,
there are many approaches, which vary in their
understanding of what is the nature of “disease”
or “problem” (pathology) and on how to resolve
or help deal with them. Most schools of therapy
derive their view of pathology and intervention
from a theory of human nature that includes,
among others, a theory of personality and devel-
opment. We hesitate to state a view of human
nature that would be acceptable to all streams of
DT. We would say that psychodynamics would
involve recognizing that people are not always
aware of the reasons for their behavior, that
human motivations are to some extent rooted
biologically, and that they are often driven by
unknown motives.
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 444
444 Chapter 12 / Research on Dynamic Therapies
Similarly, it would be wonderful if we could
easily define what DT is and what it is not.
There have been several attempts to characterize
psychodynamic psychotherapy (e.g., Blagys &
Hilsenroth, 2000); however, it would be almost
impossible to find a set of necessary and sufficient
conditions to define what is psychodynamic. Nev-
ertheless, DT can be captured by the following
characteristics: focus on unconscious processes;
focus on affect, cognitions, wishes, fantasies
and interpersonal relationships; lack of tradi-
tional homework; relatively less guidance, use of
open-ended questions; use of interpretation and
clarification; consideration of the transference
and countertransference; and use of the thera-
peutic relationship to increase self-awareness,
self-understanding, and exploration.
One of the questions we needed to address
is whether interpersonal therapy (IPT; Klerman,
Weissman, Rounsaville, & Chevron, 1984), pop-
ular in the treatment of depression, is a form
of DT. IPT is similar to many DTs because
it focuses on underlying schemas and repeti-
tive scenarios involving loss and transition, and
uses empathy, exploration of painful affects, and
(rarely) transference interpretations. It differs in
that it is highly focused and includes a signif-
icant educational component. In fact, there is
research to suggest that in practice IPT is more
like CBT than DT (e.g., Ablon & Jones, 2002).
Our opinion is that depending on the training of
the IPT therapists (e.g., their previous exposure
to DT or the extent of their IPT training), the
differences between IPT and DT can be elusive.
Some meta-analyses of DT (e.g., Leichsenring
2001) included studies of IPT and showed that
the overall meta-analytic results were unchanged
if excluded. To be conservative, we decided not to
include IPT as a form of DT therapy, mainly due
to its explicit rejection of some critical dynamic
In this chapter we cover three general aspects
of DT: its efficacy for symptom improvement;
the therapeutic alliance and its role in DT; and
processes of change, including mechanisms and
outcome unique to DT and the correlates of DT
In the past two decades, there have been increas-
ingly more studies examining the therapeutic
efficacy of DT (e.g., Shedler, 2010).1However,
it seems that the number of meta-analyses has
increased even more. The majority of these
meta-analyses have focused on short-term DT
(STDP) rather than long-term DT (LTDP) or
psychoanalysis (for notable exceptions concern-
ing possible LTDP superiority over controls
and less intense treatments, see Leichsenring &
Rabung, 2008, 2011; Smit et al., 2012). Briefly
speaking, STDP differs from LTDP in that it
generally: (a) takes place in a shorter, often time-
limited format (usually at least 8 sessions, often
12 to 20, but sometimes up to 40 depending on
the manual, disorder, etc.); (b) typically targets
specific symptomatic versus global or structural
change; and (c) does so by often identifying and
working through a relatively specific dynamic-
interpersonal focus conceptualized to underlie
expressed psychopathology (e.g., a patient’s core-
conflictual relationship theme [CCRT] in
Luborsky’s [1984] supportive-expressive ther-
apy [SE]).
Interventions in STDP are often classified
as being on a continuum of “expressive” to “sup-
portive” (Luborsky, 1984). Broadly, expressive (or
interpretive) techniques such as transference and
defense interpretations seek to augment patient
insight and understanding of in-the-moment or
repetitive interpersonal and intrapsychic pat-
terns or conflicts. Insight or self-understanding
may allow a patient to better tolerate distressing
thoughts, affects, and fantasies that are defended
against in a way that creates psychopathology;
insight might also potentiate working through
repetitive intrapsychic and interpersonal dynam-
ics previously disavowed, misunderstood, or
unknown. Supportive techniques are conceived
both in terms of building the therapeutic alliance
and of boosting client capacity to use extant
healthy capabilities (e.g., adaptive defenses, social
networks) when such a capacity may be compro-
mised (see later in the chapter for a more extensive
discussion on both the alliance and dynamic
1Due to space restrictions, our original chapter was
sharply shortened as we approached our publication
deadline. In particular, we had to cut our reference list
by more than 50%. We also removed several additional
meta-analyses that we had conducted in preparation for
this chapter. For space reasons we were required to
eliminate references of many studies and meta-analyses,
of measures used, and of many theoretical manuals
underpinning delivered treatments. We apologize in
advance to researchers whose work we are not citing.
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 445
Research on Addressing the Efficacy of DT 445
interventions). Expressive and supportive inter-
ventions are often thought to work synergistically
to promote patient change. For example, support-
ive interventions may create a positive “holding
environment involving therapist and patient that
serves as a uniquely safe interpersonal space for:
(a) experiments in self-change (e.g., trying out
diferent modes of relating through the therapist-
patient interaction); (b) expressive exploration of
thoughts, feelings, and fantasies; and (c) experi-
encing emotions and insights in an impactful way
as a result of reduced defensivenss.
Generally, previous meta-analyses of DT
reached one or more of the following four differ-
ent conclusions: (1) DT outcome does not differ
from alternative therapies (e.g., for DT across
disorders, see Leichsenring, Rabung, & Leibing,
2004; for depression, see Leichsenring, 2001; for
personality disorders, see Leichsenring & Leib-
ing, 2003); (2) there are small ES differences in
favor of DT (e.g., Anderson & Lambert, 1995, at
follow-up in some analyses of DT); (3) there are
small differences in favor of alternative treatments
across disorders (e.g., Tolin, 2010, as compared
to some CBTs, though including controls as DT
treatments) or in depression specifically (e.g.,
Driessen et al., 2010 finding a small difference
at termination but not follow-up); and (4) DT is
significantly superior to control treatments (e.g.,
for DT across disorders, see Abbass, Hancock,
Henderson, & Kisely, 2006; Leichsenring et al.,
2004; for somatic conditions specifically, see
Abbass, Kisely, & Kroenke, 2009).
In a representative and rigorous meta-
analysis of patients with different disorders,
Leichsenring et al. (2004) identified 17 well-
conducted post-1970 studies of STDP, and
found that STDPs yielded large pretreatment-
posttreatment ESs for target problems (d=1.39),
general psychiatric symptoms (d=.90), and social
functioning (d=.80). The ESs of STDP sig-
nificantly exceeded those of waiting-list controls
(d=.27) and treatments as usual (d=.55) for
treatment of target problems. No significant
differences were found between STDP and other
forms of psychotherapy at either termination or
Meta-analyses have also differed in their
inclusion criteria; some of them included any
study that mentioned DT while others were
more selective. For example, a recent Cochrane
review of STDP by Abbass et al. (2006) found
57 studies of STDP, but excluded 34 studies
for various design issues, mostly because they
did not have a control treatment group. They
also excluded studies that had more than a 20%
dropout rate. Nevertheless, their meta-analysis
included 23 RCTs comprising 1,431 patients,
and found moderate-to-large between-groups
ES advantages for STDP over control groups on
measures of general psychopathology, anxiety,
and depression at termination, short-term and
long-term follow-up.
Keeping in mind that there are few high-
quality studies of DT (Gerber et al., 2011; see
also Thoma et al., 2012, for a similar finding
in the CBT literature), we considered different
inclusion criteria with the wish to include only
studies that had the potential to be high quality.
We considered strictly following the criteria
used by the task force of Division 12 of the
American Psychological Association for desig-
nating psychotherapies as empirically supported
(Task Force on Promotion and Dissemination of
Psychological Procedures, 1995; Chambless &
Hollon, 1998), as the most convincing evidence
for the scientific world and for governmental
agencies comes from RCTs. We recognized that
these criteria are controversial (e.g., Westen,
Novonty, & Thompson-Brenner, 2004) and that
RCTs have unique drawbacks in psychother-
apy research. Those drawbacks may be more
pronounced for the study of DT, which may
for example require on the part of the patient a
certain amount of psychological mindedness and
willingness to introspect to have a chance ofwork-
ing (cf. Barber, 2009). Because manualized DTs
seldom include session-by-session guidelines, we
included studies where manualization was not
definite and to test whether manualization makes
a difference.
Methodology for the
The inconsistency of results across existing
meta-analyses could have been the result of
inconsistencies on whether the analysis focused
on a specific disorder or not, or whether they
included only RCTs or any outcome study. As a
way to address those issues and to be consistent
with recent trends, we focused on addressing
the question of whether DT is effective for
the treatment of specific disorders or disorder
groups. This resulted in a series of separate
meta-analyses for the following disorder groups:
depression, anxiety disorders, and personality
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 446
446 Chapter 12 / Research on Dynamic Therapies
Study Inclusion Criteria
The primary sample of DT studies reviewed
and meta-analyzed was taken from the Gerber
et al. (2011) quality-based review of RCTs of
DT. We used the same terms to search for more
recent publications. In addition, we reexamined
past reviews and meta-analyses for studies missed
in the aforementioned searches and contacted
experts in the field for citations for recently
completed studies.
To be included, a study had to be an RCT
comparing individual DT for adults to either
a control condition (e.g., TAU, wait list) or
alternative non-DT intervention (e.g., CBT,
pharmacotherapy). However, we did include
mentalization-based treatment (MBT; Bate-
man & Fonagy, 2006) which includes group
therapyas it is a state-of-the-art, well-defined
DT for borderline personality disorder and does
include a strong individual component. For sim-
ilar reasons, we included dialectical-behavioral
therapy (DBT) as a comparison treatment despite
it consisting of both individual and group ther-
apies. We excluded studies that had an eclectic
treatment incorporating dynamic ideas. We also
did not include studies comparing combined DT
and pharmacotherapy (e.g., SSRI) to another
combined group using a different therapy (e.g.,
Statistical Analyses
Calculations of weighted mean effect size (ES),
heterogeneity, and moderator analyses were
conducted using Comprehensive Meta-Analysis,
version 2.2.046 (Borenstein et al., 2005); a priori,
it was decided to conduct our meta-analyses using
a random-effects model for a more stringent
and generalizable test of the efficacy of DT. For
our ES measure, we used Hedge’s gsimilar
to the traditional Cohen’s d, but corrected for
upward bias (Hedges, 1981). Both gand dare
ES statistics standardized by standard deviations
units, wherein a g/dof 1.00 indicates that the
difference between two means is one (pooled)
standard deviation unit large. Values of 0.20,
0.50, and 0.80 are considered small, medium,
and large effects, respectively, in social science
research, although these conventions apply only
to between-group ESs (Cohen, 1992).
Between-groups ESs using outcome scores
on continuous measures were meta-analyzed.
In the event of incomplete reporting of scores,
ESs were imputed or estimated from available
data and reported statistical tests as per Lipsey
and Wilson (2001). For overall analyses, ESs
calculated from scores were chosen over ESs
calculated from binary outcomes (e.g., remission)
if available, except as noted. When relevant data
or tests were not available, attempts were made to
contact corresponding authors.2Meta-analyses
were performed for outcomes at termination,
and when at least three relevant studies with
data were available—at short-term follow-up
(3 to 9 months posttermination) and long-term
follow-up (> 9 months posttermination) periods.
So as to not violate independence when com-
paring DT to multiple active treatments within
the same study (e.g., behavioral therapy and cog-
nitive therapy in Gallagher & Thompson, 1982),
data from active treatment groups were collapsed
as per formulas in the Cochrane Handbook
for Systematic Reviews of Interventions (2011).
Similarly, when multiple bona fide DTswere per-
formed within the same study (e.g., transference-
focused therapy and manualized dynamic-suppor-
tive therapy in Clarkin, Levy, Lenzenweger, &
Kernberg, 2007), data from these groups were
collapsed using the same methodology.
However, for one meta-analysis (depres-
sion), we had a sample of studies that allowed
us to explore whether DT was differently effi-
cacious from medication and/or alternative
psychotherapies. Because it would be invalid to
include the same study twice, as it would violate
ES-independence criteria for meta-analysis, we
included in a particular moderation analysis only
the between-groups ES for the treatment type as
the focus of that moderation.
Heterogeneity of ESs was examined using
the Q statistic and the I2index (Higgins &
Thompson, 2002). Significant Q statistics indi-
cate that the observed range of ES is significantly
larger than what would otherwise be expected
based on within-study variances; the I2index
is a quantification of this heterogeneity, with
25%, 50%, and 75% reflecting respectively low,
medium, and high heterogeneity. Robustness
of meta-analytic results was also examined by
performing a sensitivity analysis to see if a given
result relied on the ES of a single study.
2Notably, we did receive data from the Clarkin et al.
(2007) study of dynamic and DBT treatment of BPD.
They sent us estimated pre- and post scores based on th e
hierar chical linear model s they ran t o analyze thei r three
treatment groups— as they were the best data available
for this study, we used t hesesc ore estimates in our meta-
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 447
Research on Addressing the Efficacy of DT 447
Regardless of observed heterogeneity for a
given meta-analysis, exploratory analyses were
conducted to assess for moderators of ES. Mod-
erators are categorical (e.g., manualization status)
or continuous (e.g., quality score) characteristics
of studies, used to predict outcome (i.e., ES).
Categorical moderators were assessed using an
analysis of variance (ANOVA) of mixed-effects
models for each variable hypothesized to influ-
ence the ES. Meta-regression analyses using
a maximum-likelihood model were conducted
to assess the effects of continuous moderators,
including an index of study quality scores using
the Randomized Controlled Trial Psychotherapy
Quality Rating Scale (RCT-PQRS; Kocsis et al.,
2010). When describing included studies, we
reported both total quality scores (ranging from
0 to 48) and subjective Item 25 “omnibus” study
quality rating (1 to 7; 1 being extremely poor, 7
being exceptionally good). Gerber et al. (2011)
demarcate a total score of 24 as the minimum for
an adequate quality study. If RCT-PQRS scores
were not available from Gerber et al. (2011),
studies were instead scored by two independent
graduate students who attained excellent agree-
ment (ICC =.97). With the exception of some
comparisons in our depression meta-analysis, all
moderator analyses should be considered highly
exploratory due to low (n< 10) study sample
size (see guidelines from Higgins & Green,
2011). Low study sample size may mean that
there is not enough data for the moderation
to be generalized, may not have enough power
to find an extant effect, and can lead to biased
results due to heightened effects of between-
study characteristic confounds and nonrandom
clustering of study characteristics. However, we
performed moderator analyses in the interests of
synthesizing a preliminary picture of the current
state of the literature and to perhaps direct future
research questions and analyses. Descriptions
of precise moderator findings have sometimes
been truncated for space reasons, but we always
reported on moderators if significant at least at
trend level (p< .10).
Publication bias was assessed by examination
of publication bias funnel plots and Duval and
Tweedie’s (2000) trim-and-fill procedure. When
asymmetry was evident in the funnel plot, we
applied Duval and Tweedie’s trim-and-fill pro-
cedure to provide an adjusted ES estimate that
corrects for the number and assumed location of
the missing studies.
Mood Disorders
Reflecting the variety of DTs and STDPs, studies
reporting on DT for unipolar depression have
used different forms of therapy. From a dynamic
perspective, the essential issues in depression
are loss and guilt over loss, along with low self-
esteem and failures in the attempt to restore
self-esteem (Freud, 1917). Treatment generally
involves: (a) encouraging greater activity in the
patient and instillation of hope; (b) identification
of major losses, anger, conflict over anger, guilt,
as well as the cycle of self-esteem restoration; and
(c) proactive attempt to understand vulnerability
in situations and change in response and behavior
to prevent recurrence (e.g., Busch, Rudden, &
Shapiro, 2004; Summers & Barber, 2009).
There have been two previous meta-analyses
focusing on STDP for depression. Leichsenring
(2001) meta-analyzed seven studies comparing
STDP to CBT, finding no significant differences
between them in either outcome scores or rates
of remission/response. The most recent meta-
analysis (Driessen et al., 2010) included a larger
sample of 23 studies: 10 of them were not RCTs,
and covered both individual and group therapy
for depression. Using data from 13 RCTs, unlike
Leichsenring the authors found that STDP was
more effective than control groups (d=0.80)
but less effective than alternative treatments (d=
0.35) at termination, though this difference
was not found at follow-up. Driessen et al. did
not find significant moderators apart from an
advantage of individual over group therapies
using uncontrolled ES.
In our meta-analysis, we have tended to be
inclusive rather than exclusive. However, we
excluded studies where DT was delivered in com-
bination with other treatments. We also sought
to explore moderators not investigated by pre-
vious meta-analyses such as type of alternative
treatments, impact of study quality, and addition
of pharmacotherapy. Our preferred outcome
measure (if available) was the Hamilton Rating
Scale for Depression (HRSD).
Controlled Comparisons
The comparison of DT for depression to control
conditions (two “treatment as usuals” [TAUs],
one wait list, and one pill-placebo) suggested a
moderate controlled ES (study n=4, subject
n=303, g=0.457 [0.097 to 0.818], p=0.013,
fail-safe Nof 12). Heterogeneity was found at a
trend level (Q=6.794, I2=55.846, p=0.079).
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 448
448 Chapter 12 / Research on Dynamic Therapies
Although we did not find any significant categor-
ical moderators, there was a trend indicating that
better study quality was associated with smaller
controlled ES (slope =−0.0368, total model
p=0.079; see Cuijpers, van Straten, Bohlmeijer,
Hollon, Andersson, 2010, for a similar finding
in the depression literature overall). As patient
population varied highly between studies (e.g.,
severity, duration, comorbidity), our findings
should be interpreted cautiously.
Active Treatment Comparisons
At termination, DT and alternative treatments
did not differ (study n=11, subject n=830,
g=−0.115 [0.257 to 0.026], p=.110), with
no significant heterogeneity among ESs. This
result was sensitive to the removal of Cooper
et al. (2003), which results in an estimate of a
small ES benefit to using alternative treatments
over DT (study n=10, subject n=696, g=
0.161 [0.315 to 0.007], p=0.040). Con-
versely, checking for publication bias via trim
and fill suggested the existence of unpublished
studies favoring DT, raising the ES estimate to
a still-insignificant g=−0.096. We did not find
any evidence that DT had significantly different
between-groups ESs when compared against
antidepressants versus alternative therapies (Q=
0.232, p=0.630), when compared against CBT
versus non-CBT treatments (Q=1.032, p=
0.310), or even when compared against CBT
or antidepressants versus remaining alternative
treatments (Q=0.785, p=0.376). Total quality
score (slope =−0.0100, ns ), manualization of
DT (Q=0.093, p=0.760), whether or not
the depressed population was geriatric (Q=
0.857, p=0.355), and number of sessions of DT
(slope =0.00128, ns ) were also not significant
moderators of ES. At termination, DT and alter-
native therapies for depressionincluding CBTs
and antidepressants likely effect similar overall
outcomes. Any overall difference, if extant but
undetected, is likely to be of small ES.
DT was also not significantly different from
alternative treatments at short-term follow-
up (study n=6, subject n=496, mean FU =
4.29 months, g=−0.122 [0.524 to 0.280], p=
0.553), with no evidence of publication bias or
that the result was due to including any single
study. However, unlike termination, ESs at short-
term follow-up did have significant heterogeneity
(Q=14.732, I2=66.061, p=0.012). In an
exploratory moderator analysis comparing alter-
native treatment types (Q=5.447, p=0.020),
we found that at short-term follow-up DT per-
formed significantly worse against CBTs (study
n=4, subject n=299, g=−0.380 [0.756
to 0.004], p=0.048) than when compared to
non-CBT treatments (study n=2, subject n=
57, g=0.530 [0.103 to 1.163], p=0.101). Some
dynamic therapists anticipate symptoms return
around termination, and this bounce in symptoms
short-term follow-up, we found a significant
moderation (Q=6.660, p=0.010) suggesting
that DT was inferior to alternative treatments
for geriatric (study n=2, subject n=96 g=
0.853 [1.484 to 0.222], p=0.008) but not
nongeriatric patient populations (study n=4,
subject n=400, g=0.086 [0.247 to 0.419],
p=0.611). It is possible that older patients may
benefit more from concrete, direct interventions
for depression because the concerns leading to
their depression may be more grounded in actual
experience associated with aging (e.g., loss of sup-
port, declining physical health) than in internal
conflict. Because all geriatric studies used CBTs
as their comparison therapies, it is difficult to
disambiguate whether the geriatric patient popu-
lation or use of CBTs (or perhaps both or neither)
was the primary reason for DT diminished per-
formance among these studies. However, in
both DT and CBT, geriatric depressed patients
improve less than other patients. Among only
studies in which DT was compared to a CBT,
DT performed significantly worse than CBTs
(Q =4.922, p=0.027) in geriatric studies (g=
0.625 and 1.096) than in nongeriatric studies
(g=−0.157 and 0.149). This finding is consis-
tent with the interpretation that the moderation
showing diminished comparative efficacy for DT
against CBTs may in fact be driven by diminished
DT efficacy for geriatric patients rather than a
superiority of CBTs per se.
Finally, DT did not differ from alternative
treatments at long-term follow-up (study n=5,
subject n=487, mean FU =23 months, g=
0.205 [0.546 to 0.136], p=0.239), again with
no evidence of publication bias or single-study
sensitivity. As at short-term follow-up, ESs at
long-term follow-up displayed significant hetero-
geneity (Q=9.772, I2=59.068, p=0.044). No
moderators were significant, notably including
geriatric patient population. Furthermore, at
long-term follow-up, all alternative treatments
investigated were CBTs (with the exception of
the supportive therapy group from Cooper et al.
[2003], which does not differ notably in ES from
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 449
Research on Addressing the Efficacy of DT 449
the study’s CBT group). Thus, the use of CBT vs.
non-CBT comparison therapies cannot explain
observed ES heterogeneity at this time point.
Combined DT +Pharmacotherapy
Versus Pharmacotherapy
Using remission rates as the common outcome
variable, we found a moderate ES advantage at
termination (study n=3, subject n=295, g=
0.470 [0.036 to 0.904], p=0.034) of combined
DT and antidepressant treatment over antide-
pressants alone, with no indication of publication
bias (see Cuijpers, Dekker, Hollon, & Andersson,
2009; de Maat et al., 2008 for similar findings).
There was, however, significant heterogeneity
among the three studies (Q=6.649, I2=69.919,
p=0.036). No moderators were significant.
Conclusions: The Efficacy of DT
for Depression
Despite limiting our meta-analysis to only RCTs
and individual DT, our results converge with
those of Driessen et al. (2010). We found that
DTs are as therapeutically effective as alternative
treatments (both psychosocial and pharmacolog-
ical) in the treatment of depression at the end
of active treatment and at short- and long-term
follow-up. DT was also found to be more effec-
tive than control conditions. We further found
that combined DT and pharmacotherapy may be
more efficacious than pharmacotherapy alone, a
finding that has not often been confirmed in the
Some versions of DTs for depression may
meet criteria for being probably efficacious as
defined by APA Division 12 “empirically sup-
ported treatments” criteria (e.g., brief dynamic
interpersonal therapy, parent-infant psychody-
namic therapy [Cooper et al., 2003]). Conversely,
no one specific version of DT for depression
meets full, formal APA Division 12 criteria to
be designated a well-established EST because
there has not been a replication by a separate
research group of any effective, manualized form
of DT for depression (Chambless & Hollon,
1998; Connolly Gibbons, Crits-Christoph &
Hearon, 2008). However, it is not always clear
just how different from each other various ver-
sions of DT for depression actually are in theory
and practice—in many instances, the manuals
appear largely convergent. This raises the inter-
esting question as to how to decide that a specific
treatment has been replicated in order to meet
Division 12 criteria. Regardless of DT’s official
status as an EST or not, readers need to keep in
mind that few treatments have been shown to be
more effective than other active treatments for
depression (Cuijpers, van Straten, Andersson, &
van Oppen, 2008). In light of the strong trend
in the meta-analytic literature of equivalence for
depression between major therapeutic models,
future investigations might pay specific attention
to which specific depressed patients might be best
suited for DT versus other therapies (see Barber
& Muenz, 1996; Barber et al., 2012).
Anxiety Disorders
Though DT is widely used to treat anxiety disor-
ders (e.g., Goisman, Warshaw, & Keller, 1999),
the empirical literature supporting its effective-
ness is nascent but growing (see Slavin-Mulford &
Hilsenroth, 2012, for a review including natural-
istic and quasi-experimental empirical evidence).
The question of whether DT is efficacious for
the treatment of anxiety disorder is one of the
most important areas of research in terms of
addressing the specific versus common hypoth-
esis of psychological change (cf. Castonguay &
Grosse Holtfort, 2005; Schut & Castonguay,
2001). Many CBT researchers may agree that
nonspecific (common) factors could explain
positive results in depression, but they would
argue that the specific aspects of CBT for anxiety
disorders are responsible for the greater rate of
improvement for those disorders (cf. Chambless,
2002; Chambless & Ollendick, 2001).
Dynamic conceptualizations of the etiology,
maintenance, and treatment of anxiety symptoms
do often differ dramatica lly from those commonly
described in the CBT traditions, and propose dif-
ferent specific mechanisms of change. DT for
anxiety disorder is well-exemplified by manual-
ized panic-focused psychodynamic psychotherapy
(PFPP; Busch, Milrod, Singer, & Aronson, 2011).
PFPP posits that panic attacks arise from specific
unconscious conflicts, most commonly conflicts
of dependency and attachment (e.g., panic as a
way to indirectly express a need for care without
requiring direct acknowledgment of surrounding
conflicts) or feelings of anger (e.g., panic as an
aggressive means of coercing attention or as a
fearful reaction to the anger felt toward essen-
tial figures). Guilt about dependent and angry
wishes is also frequently a contributor, with panic
acting as a means of unconscious punishment.
Treatment involves interpreting the emotional
significance of panic, the psychological meaning
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 450
450 Chapter 12 / Research on Dynamic Therapies
of specific symptoms and panic triggers, identify-
ing the relevant intrapsychic conflicts alluded to
above, and increasing the understanding of inner
experiences related to panic and their underlying
dynamics. As in many DTs, this would entail
working through repetitive conflicts surrounding
the psychological meanings of panic through
demonstration that the same conflicts may be
emerging in multiple settings, including in the
To date there has never been a meta-analysis
synthesizing the literature on the effects of DT in
the control led treatment of anxiety disorders . Due
to the paucity of studies, we have decided to focus
on anxiety disorders in general. So as to include
multiple anxiety disorders in the same meta-
analysis, we used the author-identified primary
outcome measure when available to calculate
ESs. When primary outcome was not explicitly
indicated, the judgment of the meta-analysis
authors was used to select outcome measures for
analysis based on the specific psychopathology of
the disorder in question. Regarding the construct
of general anxiety, the clinician-based Hamil-
ton Rating Scale for Anxiety (HRSA) was our
preferred outcome measure.
Controlled Comparisons
Keeping in mind the very small number of studies
analyzed, DT was superior to control conditions
with a large estimated controlled ES (study n=
3, subject n=105, g=0.775 [0.381 to 1.168], p<
0.001, fail-safe N=10). Two studies used a
minimal treatment control (Alstrom et al., 1984a,
1984b; providing study assessments, psychoe-
ducation, and recommendation/instruction for
self-exposure but no ongoing treatment), while
one used a wait-list control (Brom, Kleber, &
Defares, 1989). There was no significant hetero-
geneity among ESs, nor was there single-study
sensitivity. A trim-and-fill check for publication
bias suggested adjustment downward to g=
0.630 (0.313 to 0.946). Given the quality of those
studies, their date of publication (none after
1990), and their small sample sizes, replication of
controlled ES for DT against controls is needed,
especially against active controls.
Active Treatment Comparisons
DT for anxiety disorders did not have signifi-
cantly different outcomes compared to alternative
treatments (study n=8, subject n=390, g=
0.083 [0.247 to 0.413], p=0.622). This result
was insensitive to the removal of any single study
from the meta-analysis, and there was no indi-
cation of publication bias. A significant level of
heterogeneity among study outcomes was found
(Q=17.544, I2=60.101, p=0.014). As sev-
eral DT treatment protocols contained explicit
references to minor exposure elements (e.g., a
recommendation to self-expose and then discuss
in SE therapy for GAD), we were curious as to
whether the inclusion of such a stipulation in
the treatment protocol moderated ESs. How-
ever, whether a DT used such a minor exposure
element (Alstrom et al., 1984a, 1984b; Crits-
Christoph et al., 2005; Leichsenring et al., 2009)
did not significantly moderate ESs (Q=0.084,
p=0.772). It remains possible, however, that in
commonly delivered DT there are elements of
exposure (cf. Lambert & Ogles, 2004). It is also
possible that commonly delivered CBTs include
elements of DT such as interpretations of uncon-
scious wishes and defenses. A single moderator
was significant: DT for disorders other than GAD
was more comparatively efficacious than DT for
GAD (Q=5.370, p=0.020). More specifically,
we found a small ES benefit to using DT over
alternative treatments when treating anxiety
disorders other than GAD (study n=4, subject
n=203, g=0.364 [0.008 to 0.720], p=0.045),
whereas DT for GAD was not significantly dif-
ferent from alternative treatments (study n=4,
subject n=187, g=−0.238 [0.602 to 0.126],
At short-term follow-up, equivalence bet-
ween DT and alternative treatments was main-
tained (study n=6, subject n=304, average
FU =6.16 months, g=−0.154 [0.691 to
0.383], p=0.573), a finding that was also insen-
sitive to removing any single study. This result
may need to be interpreted cautiously, as a high
level of heterogeneity was found among ESs
(Q=25.087, I2=80.070, p< 0.001). In addi-
tion, a trim-and-fill check for publication bias
imputed the existence of studies to the right
of the estimated ES, raising the estimate to an
insignificant g =0.182. Using exploratory mod-
erator analysis we found some suggestion that,
by follow-up, DT for GAD may be significantly
less effective than alternative treatments (Q=
5.310, p=0.021). At short-term follow-up, DT
for GAD was moderately worse than alternative
treatments (study n=3, subject n=159, g=
0.601 [1.125 to 0.077], p=0.025), while
there was no significant difference between DT
and other therapies when treating other anxiety
disorders (study n=3, subject n=145, g=0.275
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 451
Research on Addressing the Efficacy of DT 451
[0.255 to 0.805], p=0.309). This follow-up
result, however, was dependent on our inclusion
of Durham et al. (1994), which was unbalanced
to the detriment of DT (i.e., no manual or even
theoretical formulation was used for DT but
were for the CBT conditions). Replication using
another manualized DT would help elucidate
to what extent our moderator analysis reflects a
“true” disadvantage of DT for GAD, or merely
bias due to differences in treatment fidelity. No
other moderator was found to be significant.
Conclusions: The Efficacy of DT
for Anxiety Disorders
Despite common beliefs to the contrary (cf.
Chambless, 2002; Tolin, 2010), the evidence
from RCTs suggests that DT is largely neither
better nor worse for treatment of anxiety disorders
than are other active treatments (predominantly
CBTs), a finding carried into short-term follow-
up. We also found evidence that DT may have
a medium-to-large ES advantage over minimal-
to-no treatment controls on primary outcome,
although among a small number of older studies.
One limitation to our meta-analysis is that
these results are —with the exception of the GAD
moderation analysis —collapsed across disorders
(see Table 12.2), and notably included no con-
trolled studies of obsessive-compulsive disorder.
With the current studies available, it is not possi-
ble to further disambiguate the relative effects of
DT for different anxiety disorders. Another lim-
itation is the low quality of some studies included
in the meta-analysis (e.g., both Alstrom studies;
Pierloot & Vinck, 1978), though study quality
did not moderate ESs. Furthermore, several
studies had pilot-level sample sizes. Nevertheless,
our meta-analysis introduces the possibility that
DT for anxiety disorders may be as efficacious
as alternative treatments. Several large scale
DT RCTs are currently underway (investigating
panic disorder and social phobia), and their results
will be published over the next few years. Future
research into the effects of DT for specific anxiety
disorders is warranted, especially for GAD.
Personality Disorders (PD)
It is always difficult to provide a summary state-
ment about the essence of DT for any disorder,
but even more so for PD due to the diversity of
PDs and of forms of DT. Following Magnavita
(1997), one could say that DT for PD generally
entails the identification of maladaptive, recurring
patterns of thinking, perceiving, and behavioral
and emotional responding, and the restructur-
ing of these patterns, primarily through linking
the current and transference patterns to early
relational disruptions of attachment and trauma.
For example, Kernberg’s transference-focused
psychotherapy (TFP) for borderline personality
disorder (BPD) proposes that the symptomatol-
ogy of BPD (e.g., ip-flopping “split” perceptions
of relationships as being all-good or all-bad,
intolerable feelings of emptiness) emerges from
pervasively unintegrated self-object represen-
tations (Clarkin et al., 2006). Accordingly, the
primary mechanism of treatment in TFP is the
gradual elucidation and integration of these split-
off self-object representations as they emerge in
the transference. In an attempt to integrate these
disparate and polarized self-object representa-
tions, the therapist brings these representations
into conscious conflict with another when other-
wise they would pass without recognition of their
mutual incoherence. Therapists further seek to
analyze both what triggers the emergence of par-
ticular representations and what defensive roles
the separation and switching between “good”
and “bad” representations serves. This is thought
to help the patient improve their reality testing,
ability to mentalize the thoughts and feelings of
others (see Bateman & Fonagy, 2006), interper-
sonal stability, and sense of internal coherence
and wholeness.
Prior synthesis of the clinical literature via
meta-analysis indicates that psychotherapyis gen-
erally efficacious in the treatment of PD (Perry,
Banon, & Ianni, 1999). Leichsenring and Leib-
ing’s (2003) meta-analysis was the first to examine
the specific efficacy of DT in the treatment of
PD, as compared to the efficacy of CBT. Based
on 15 studies (6 controlled, 9 naturalistic), they
reported large within-groups (or uncontrolled)
ESs for both overall change (d=1.46) and spe-
cific measures of personality pathology (d=1.56
from the subset of 6 studies that reported person-
ality pathology scores). The uncontrolled ES for
DT was not significantly different than the one
calculated for CBT (d=1.00 from a sample of ten
studies). Notably, all ESs were calculated using
data from the furthest point available from termi-
nation (average 78 weeks for DT and 13 weeks for
CBT), perhaps explaining why the ES for DT was
arithmetically (but not significantly) larger than
the one found for CBT. From those results, they
inferred that DT (and, to a lesser extent, CBT)
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 452
452 Chapter 12 / Research on Dynamic Therapies
can cause long-term, sustained change in psy-
chopathology. More recently, Abbass, Town, and
Driessen (2011) meta-analyzed a small number
of controlled and uncontrolled studies examining
DT for patients with depression and co-morbid
personality pathology, finding both large within-
group ESs and no significant differences in effi-
cacy compared to other investigated treatments.
In the current meta-analysis, we focused on
between-groups ES estimates for PD treatment,
averaging together ESs from patient-reported
and/or observer-reported measures of psy-
chopathology within the study. We refer to this
composite ES as “general outcome.” Assessing
general outcome across measures allows for
the inclusion of more studies within the meta-
analysis. However, there is broad disagreement
as to what precisely constitutes change within a
given PD, never mind across PDs. As such, we
also performed a secondary set of meta-analyses
for both between-groups and controlled ESs
for improvement in specific psychopathology
construct for which at least three studies could
contribute an ES (e.g., three controlled studies
used a variant of the Inventory for Interper-
sonal Problems [IIP; Horowitz, Rosenberg, Baer,
Ureño, & Villaseñor, 1988]).
Controlled Comparisons
At termination, DT for PD was more effective
than controls for general outcome (study n=
7, subject n=452, g=0.593 [0.258 to 0.918],
p=0.001, fail-safe N=52). Control condi-
tions included two enhanced TAUs (Bateman &
Fonagy, 2009; Doering et al., 2010), two TAUs
(Bateman & Fonagy, 1999; Gregory et al., 2008),
and three wait-list controls (Abbass, Sheldon,
Gyra, & Kalpin, 2008; Emmelkamp et al., 2006;
Winston et al., 1994). Though this comparison
showed moderate heterogeneity (Q=15.903,
I2=62.272, p=0.014) — as to be expec ted
from including different intensities of control
treatments there was no evidence of publica-
tion bias or that the result was driven by any
single study.
The heterogeneity of control conditions
used should caution against overinterpretation
of the precise magnitude of the controlled effect
size. However, we could not meaningfully test
moderator effects of using active controls versus
inactive controls or testing for borderline per-
sonality disorder (BPD) patients versus other PD
patients because all BPD studies used an active
control and all others used an inactive control.
Furthermore, we could not examine the effects of
manualization as only one study in this compari-
son (Emmelkamp et al., 2006) did not qualify as
manualized; this studycomparing DT versus
a wait-list for avoidant PD treatment had the
lowest controlled ES (g=−0.030). No modera-
tors were significant, though future explorations
of moderators (e.g., number of therapy sessions)
might be better served using samples with the
same types of control treatments (e.g., only versus
Active Treatment Comparisons
DT did not differ from alternative therapies for
PD in terms of general outcome (study n=7,
subject n=528, g=−0.145 [0.342 to 0.052],
p=0.150). This ES estimate was not impacted
by significant heterogeneity or indication of
publication bias, nor did any one study drive this
finding. As only one study was unmanualized
(Emmelkamp et al., 2006; though see footnote 4),
it was not possible to investigate whether man-
ualization affected ES. No moderators were
significant. In particular, we did not find signifi-
cant moderation for whether the study concerned
the treatment of BPD (Q=0.093, p=0.761) or
a primary Cluster-C disorder. (Q=0.152, p=
In terms of short-term follow-up, there was
again no significant difference in general outcome
between DT and alternative treatments (study
n=5, subject n=381, average FU =6months,
g=−0.056 [0.367 to 0.255], p=0.723). There
was a trend toward heterogeneity among ESs at
follow-up (Q=7.784, p< 0.100, I2=48.616),
though there was no indication of publication
bias. The result was not single-study sensitive.
Again, only Emmelkamp et al. (2006) was not
manualized, though it did have the lowest ES
(g=−0.658). Exploratory meta-regression sug-
gested a trend of a significant, positive rela-
tion between number of dynamic sessions and
between-groups ES (slope =0.04395, total model
p< 0.100).
Secondary Analyses: Specific Treatment
Outcomes of DT for PD
In a secondary set of analyses, we performed
meta-analyses for both between-groups and
controlled ESs for improvement in any specific
psychopathology construct for which at least
three studies could contribute an ES (e.g., three
controlled studies used a variant of the IIP). Six
constructs were identified: personality pathology
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 453
Research on Addressing the Efficacy of DT 453
(e.g., STIPO, PDQ); general symptomatology
(e.g., SCL-90); global functioning (measured
by the GAF); interpersonal problems (measured
by the IIP); depression (measured by the BDI);
and suicidality (e.g., rates of patients attempting
suicide). All six constructs had enough stud-
ies contributing data to perform preliminary
meta-analyses for controlled outcome. However,
only three constructs (general symptomatology,
personality pathology, and interpersonal prob-
lems) had enough studies to meta-analyze DT
compared to other active therapies.
We found no significant differences between
DT and other therapies on measures of personal-
ity pathology (study n=6, g=−0.108 [0.357
to 0.140], p=0.392), general symptomatology
(study n=4, g=−0.078 [0.291 to 0.136], p=
0.476), and interpersonal problems (study n=
4, g=0.019 [0.194 to 0.232], p=0.861). For
controlled ES, DT had a significant advantage
over control treatments on measures of general
symptomatology (study n=5, g=0.565 [0.135
to 0.994], p=0.010), suicidality (study n=4, g=
0.649 [0.394 to 0.904], p=0.000), global func-
tioning (study n=3, g=0.579 [0.204 to 0.955],
p=0.002), interpersonal problems (n=3, g=
1.245 [0.463 to 2.028], p=0.002), personality
pathology (n=3, g=0.311 [0.015 to 0.607], p=
0.040), and a trend toward an advantage in treat-
ing depressive symptomatology (study n=4, g=
0.645 [0.060 to 1.349], p=0.073). It is possible
that the controlled ES for personality pathology
outcome may be somewhat underestimated at
treatment termination relative to follow up, as
both Bateman and Fonagy (1999 for original
study, 2008 for follow-up report) and Gregory
et al. (2008 for original, 2010 for follow-up)
report long-term follow-up between-groups ESs
for personality pathology that are substantially
larger than our estimate here (d=1.80 and d=
1.31, respectively).
Conclusions: The Efficacy of DT for PD
Among the small number of RCTs of DT for
PDs, DT is superior to control conditions,
but not different than alternative treatments in
terms of general psychopathological outcome
at termination and short-term FU. We did not
find any indication from exploratory moderator
analysis that DT showed significantly different
between-groups ESs when treating either BPD or
Cluster-C PDs. It may behoove future dynamic
trials of PD treatments to use more active con-
trols to better assess the efficacy of DT treatment
and examine whether benefits of DT over active
controls is maintained after termination (for evi-
dence that this may be true, see follow-up reports
Bateman & Fonagy, 2008; Gregory et al., 2010).
Given the nature of DT (its goals and pro-
cesses), there may be the expectation that DT
would do well with and perhaps better than other
types of psychotherapy with personality pathology
and interpersonal relations. However, we found
that overall equivalence between DT and other
treatments for PD is the rule when specifically
examining personality pathology, general symp-
tomatology, and interpersonal problems. It seems
that the Dodo Bird verdict (Luborsky, Rosenthal,
et al., 2002) applies to existing controlled studies
of PD up to short-term follow-up. We further
showed superiority of DT over controls in the
domains of general symptomatology, suicidal-
ity, global functioning, interpersonal problems,
personality pathology, and—less definitively
depressive symptoms. With regard to controlled
effect, DT may have the strongest evidence base
in the domain of PD treatment.
In fact, DT’s first full “EST” has come from
the treatment of personality pathology: TFP
has recently been designated a well-established
treatment for BPD by APA Division 12 (2012).
Mentalization-based treatment has also been
determined to be probably efficacious for BPD
(APA Division 12, 2012) for exhibiting clear
superiority to both a strong TAU and a manu-
alized enhanced TAU in two RCTs, one RCT
showing large ES advantages over TAU over 5
years after treatment. Several individual manu-
alized variants of DT might also be considered
at least probably efficacious for the treatment of
PDs as per APA criteria (Chambless & Hollon,
1998): Vaillant McCullough’s STDP (Svartberg
et al., 2004 for Cluster-C), brief relational ther-
apy (Muran et al., 2005 for Cluster C), intensive
STDP (Abbass et al., 2008; Winston et al., 1994
for general PD), brief adaptive psychotherapy
(Muran et al., 2005, Winston et al., 1994 for
general PD and Cluster C), psychodynamic psy-
chiatric management (McMain et al., 2009/2012,
for BPD), and manualized dynamic-supportive
therapy (Clarkin et al., 2007, for BPD, though
less preferable to TFP on measures of suicidality).
Unambiguously, DTs should be considered
viable and efficacious treatments for personality
pathology. Especially in consideration of the
dearth of well-studied, efficacious treatments for
PDs compared to other disorders, replications
and extensions of several DT treatments for PDs
are highly warranted. Clearly, there is a need to
Lambert c12.tex V2 - 11/27/2012 5:32pm Page 454
TABLE 12.1 Description of DT RCTs of Mood Disorders (Depressive Dxs)
Period (mos.)
Quality Score
Barber et al. (2012) Chronic MDD 20 sessions SE,
MAN (n=51)
ADP (n=55),
PL (n=50)
None SE =ADP =PL 43/6 HRSD
Barkham et al. (1999) Subsyndromal,
mild MDD
2 weekly sessions
then 1 session 3
months later
Long-term FU:
29/4 BDI
Burnand, Andreoli,
Kolatte, Venturini,
and Rosset (2002)
MDD 82 inpatient days of
UNMAN (n=35)
ADP (n=39) None DIT +ADM > ADM 27/4 HRSD
Cooper, Murray,
Wilson, and
Romaniuk (2003)
Postpartum 10 sessions PIPT,
MAN (n=45)
CBT (n=42),
NDC (n=47),
TAU (n=50)
4, 13, and 55 Termination: PIPT =
CBT, NDC on scores
but PIPT > CBT,
NDC on remission;
Long-term FU:
37/5 EPDS
de Jonghe, Kool, van
Aalst, Dekker, and
Peen (2001)
MDD with or
8 weekly and 8 sessions
every 2 weeks DST
ADP (n=57) None DST +AD > AD 35/6 HRSD
Gallagher and
Thompson (1982)
Geriatric MDD 16 sessions over
12 weeks BRT,
UNMAN (n=10)
CT (n=10),
BT (n=10)
1.5, 3, 6,
and 12
Termination: BRT =
Short- and long-term
FU: maintained
27/4 HRSD
and Steffen (1994)
Geriatric major,
minor, or
16–20 sessions BDT,
UNMAN (n=30)
CBT (n=36) 3 BDT =CBT 27/6 RDC
or remission
Hersen, Bellack,
Himmelhoch, and
Thase (1984)
MDD 12 sessions BDT +PL