ArticlePDF Available

All just a Question of the larger dose? Why analytic Psychotherapy in depressive Disorders is effective in the long term

January 2014
Zeitschrift fur Psychosomatische Medizin und Psychotherapie 60(1):70-71

January 2014
60(1):70-71

Authors:

Medical University of Vienna

Empirical evidence for the effectiveness of long-term psychodynamic psychotherapy (LTPP) in patients with mood disorders is growing. However, it is unclear whether the effectiveness of LTPP is due to distinctive features of psychodynamic/psychoanalytic techniques or to a higher number of sessions. We tested these rival hypotheses in a quasi-experimental study comparing psychoanalytic therapy (i.e., high-dose LTPP) with psychodynamic therapy (i.e., low-dose LTPP) and cognitive–behavioural therapy (CBT) for depression. Analyses were based on a subsample of 77 subjects, with 27 receiving psychoanalytic therapy, 26 receiving psychodynamic therapy and 24 receiving CBT. Depressive symptoms, interpersonal problems and introject affiliation were assessed prior to treatment, after treatment and at the 1-, 2- and 3-year follow-ups. Psychoanalytic techniques were assessed from three audiotaped middle sessions per treatment using the Psychotherapy Process Q-Set. Subjects receiving psychoanalytic therapy reported having fewer interpersonal problems, treated themselves in a more affiliative way directly after treatment and tended to improve in depressive symptoms and interpersonal problems during follow-up as compared with patients receiving psychodynamic therapy and/or CBT. Multilevel mediation analyses suggested that post-treatment differences in interpersonal problems and introject affiliation were mediated by the higher number of sessions, and follow-up differences in depressive symptoms were mediated by the more pronounced application of psychoanalytic techniques. We also found some evidence for indirect treatment effects via psychoanalytic techniques on changes in introject affiliation during follow-up. These results provide support for the prediction that both a high dose and the application of psycho-analytic techniques facilitate therapeutic change in patients with major depression. Copyright © 2014 John Wiley & Sons, Ltd.

Pre-treatment sample description Psychoanalytic therapy Psychodynamic therapy Cognitive–behavioural therapy

…

. Descriptive statistics and intercorrelations of target variables

…

. Predicting dose and psychoanalytic technique (Path a)

…

. Relative specific indirect effects transmitting the effect of psychoanalytic therapy to long-term outcome

…

Figures - uploaded by Henriette Loeffler-Stastka

Content may be subject to copyright.

Content uploaded by Henriette Loeffler-Stastka

Content may be subject to copyright.

Is It All about the Higher Dose? Why

Psychoanalytic Therapy Is an Effective Treatment

for Major Depression

Johannes Zimmermann,

*Henriette Löfﬂer-Stastka,

Dorothea Huber,

3,4

Günther Klug,

Sarah Alhabbo,

Astrid Bock

and Cord Benecke

Department of Psychology, University of Kassel, Kassel, Germany

Department of Psychoanalysis and Psychotherapy, Medical University of Vienna, Vienna, Austria

International Psychoanalytic University Berlin, Berlin, Germany

Department of Psychosomatic Medicine and Psychotherapy, Klinikum München-Harlaching, Munich, Germany

Department of Psychosomatic Medicine, Technical University Munich, Munich, Germany

Institute of Psychology, University of Innsbruck, Innsbruck, Austria

Empirical evidence for the effectiveness of long-term psychodynamic psychotherapy (LTPP) in patients

with mood disorders is growing. However, it is unclear whether the effectiveness of LTPP is due to

distinctive features of psychodynamic/psychoanalytic techniques or to a higher number of sessions.

We tested these rival hypotheses in a quasi-experimental study comparing psychoanalytic therapy

(i.e., high-dose LTPP) with psychodynamic therapy (i.e., low-dose LTPP) and cognitive–behavioural

therapy (CBT) for depression. Analyses were based on a subsample of 77 subjects, with 27 receiving

psychoanalytic therapy, 26 receiving psychodynamic therapy and 24 receiving CBT. Depressive

symptoms, interpersonal problems and introject afﬁliation were assessed prior to treatment, after

treatment and at the 1-, 2- and 3-year follow-ups. Psychoanalytic techniques were assessed from

three audiotaped middle sessions per treatment using the Psychotherapy Process Q-Set. Subjects

receiving psychoanalytic therapy reported having fewer interpersonal problems, treated themselves

in a more afﬁliative way directly after treatment and tended to improve in depressive symptoms

and interpersonal problems during follow-up as compared with patients receiving psychodynamic

therapy and/or CBT. Multilevel mediation analyses suggested that post-treatment differences in

interpersonal problems and introject afﬁliation were mediated by the higher number of sessions,

and follow-up differences in depressive symptoms were mediated by the more pronounced

application of psychoanalytic techniques. We also found some evidence for indirect treatment

effects via psychoanalytic techniques on changes in introject afﬁliation during follow-up. These

results provide support for the prediction that both a high dose and the application of psycho-

2014 John Wiley & Sons, Ltd.

Key Practitioner Message:

•Psychoanalytic therapy is an effective treatment for major depression, especially in the long run.

•The differential effectiveness of psychoanalytic therapy cannot be fully explained by its higher dose.

•Distinctive features of psychoanalytic technique (e.g., focusing on patients’dreams, fantasies,

sexual experiences or childhood memories) may play an important role in establishing sustained

therapeutic change.

Keywords: Long-Term Psychodynamic Psychotherapy, Psychoanalytic Technique, Dose Effect, Major

Depression, Mediators of Change, Psychoanalysis

During the last decade, empirical evidence for the effec-

tiveness of psychodynamic psychotherapies has grown

steadily (Gerber et al., 2011; Gibbons, Crits-Christoph, &

Hearon, 2008; Shedler, 2010). This is especially true for

short-term psychodynamic psychotherapy (STPP), which

has been empirically supported as an effective treatment

of speciﬁc mental disorders by several randomized con-

trolled trials (for meta-analyses on major depression, see,

*Correspondence to: Johannes Zimmermann, Department of Psy-

chology, University of Kassel, Kassel, Germany.

E-mail: johannes.zimmermann@uni-kassel.de

Clinical Psychology and Psychotherapy

Clin. Psychol. Psychother. 22, 469–487 (2015)

Published online 4 September 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpp.1917

e.g., Cuijpers, van Straten, Andersson, & van Oppen,

2008; Driessen et al., 2010; Leichsenring, 2001). There is

also emerging evidence for long-term psychodynamic

psychotherapy (LTPP), which seems to be superior to

shorter treatments both at post-treatment and at follow-

up (Leichsenring & Rabung, 2008, 2011; Leichsenring,

Abbass, et al., 2013).

In these studies, LTPP is deﬁned

by at least 50 sessions of psychodynamic treatment or by

a psychodynamic treatment that lasts for at least 1 year.

For even more intensive psychodynamic treatments (e.g.,

psychoanalytic therapy with more than 100 sessions), the

evidence is still limited, although prospective cohort

studies have reported substantial pre–post changes in pa-

tients with complex mental disorders (de Maat et al., 2013).

Despitethegrowingevidencefortheeffectivenessof

LTPP in severely disturbed patients, it is unclear which

distinctive features of such treatments are responsible for

their effectiveness. Two features are frequently invoked

as explanations in the literature: a higher number of

sessions (e.g., Smit et al., 2012) and the speciﬁc

psychodynamic/psychoanalytic techniques applied by

the therapist (e.g., Shedler, 2010). Put more pointedly, the

question is whether intensive psychodynamic treatments

are effective because they are intensive or because they are

psychodynamic/psychoanalytic. The aim of this study was

to test these rival hypotheses using data from the Munich

Psychotherapy Study (MPS; Huber, Henrich, Clarkin, &

Klug, 2013; Huber, Henrich, Gastner, & Klug, 2012;

Huber, Zimmermann, Henrich, & Klug, 2012), which is

a quasi-experimental study comparing psychoanalytic

therapy (i.e., high-dose LTPP) with psychodynamic ther-

apy (i.e., low-dose LTPP) and cognitive–behavioural

therapy (CBT) for depression.

Dose Effects on Psychotherapeutic Outcomes

The most prominent explanation for the effectiveness of

LTPP in the current literature is the higher number of ses-

sions as compared with shorter treatments. For example,

Smit et al. (2012, p. 89) noted that ‘any comparison with

STPP is (…) complicated, as these studies do not inform

us about the causes of a difference in effect size, apart

from treatment duration. For example, differences might

be purely attention and intensity effects, not related to

psychoanalytic therapy per se’. In fact, a relatively stable

ﬁnding across studies is that the amount of psychotherapy

is positively related to patient improvement, albeit in a

nonlinear fashion (Lambert & Ogles, 2004). In their

seminal meta-analysis on the dose–response relationship,

Howard, Kopta, Krause, and Orlinsky (1986) found that

the effect of dose on psychotherapeutic outcome follows

a negatively accelerated curve, with higher rates of

improvement in earlier sessions and lower rates in later

sessions. This pattern has been conﬁrmed in many

naturalistic studies involving a broad range of outcome

measures, patient characteristics, treatment types and

number of sessions (e.g., Anderson & Lambert, 2001;

Kopta, Howard, Lowry, & Beutler, 1994; Lambert,

Hansen, & Finch, 2001). Generally, these studies indicate

that, after 10 sessions, a sizeable portion of patients show

clinically signiﬁcant symptom improvements, but more

than 50 sessions are needed to reach a response rate of

75% (Lambert & Ogles, 2004). Dose effects have also been

reported for interpersonal or personality problems (Kopta

et al., 1994) and for more severely disturbed patients

(Anderson & Lambert, 2001), although the rates of

improvement were expectably lower. Moreover, it should

be noted that several studies experimentally tested the

effect of dose on psychotherapeutic outcome by randomly

assigning depressed patients to 8 or 16 sessions of

(different kinds of) psychotherapy (e.g., Barkham et al.,

1996; Dekker et al., 2005). However, the results were rather

mixed and seem to be of limited utility when estimating

dose effects in long-term treatments. In this regard,

ﬁndings from two recent meta-analyses on LTPP are more

informative: Leichsenring and Rabung (2011) found that

within-group effect sizes were positively correlated with

the number of sessions across all treatment conditions,

ranging from r= 0.37 for psychiatric symptoms to r= 0.63

for social functioning. Moreover, Smit et al. (2012) con-

ducted an exploratory meta-regression and found some

indication that the between-group effect size of LTPP

could be predicted by the ratio of sessions across groups,

with an increasing ratio (i.e., the number of LTPP sessions

divided by the number of control sessions) showing larger

effect sizes. Both ﬁndings corroborate the results from

naturalistic studies by showing that the effect of dose on

psychotherapeutic outcome can also be detected in

treatments with more than 50 sessions. In sum, there is a

great deal of evidence supporting the hypothesis that

the effectiveness of LTPP mainly stems from its high

treatment dose.

Effects of Psychodynamic/Psychoanalytic Techniques

on Psychotherapeutic Outcomes

The alternative explanation for the effectiveness of

LTPP invokes distinctive features of psychodynamic/

psychoanalytic techniques. Several authors have provided

deﬁnitions of what is essential to psychodynamic or psy-

choanalytic psychotherapy (Ablon & Jones, 1998, 2005;

Blagys & Hilsenroth, 2000; Fonagy & Kächele, 2009; Henry,

We acknowledge that some researchers question the existing evi-

dence for the effectiveness of LTPP (Smit et al., 2012). However, a de-

tailed discussion of this controversy is beyond the scope of this paper

and can be found in the work of Leichsenring, Abbass, et al. (2013).

470 J. Zimmermann et al.

Strupp, Schacht, & Gaston, 1994; Krause, 2009). For

example, in their review of the comparative psychotherapy

process literature, Blagys and Hilsenroth (2000) delineated

seven interventions that distinguished psychodynamic–

interpersonal therapy from CBT: (a) a focus on affect and

the expression of patients’emotions; (b) an exploration of

patients’attempts to avoid topics or engage in activities that

hinder the progress of therapy; (c) the identiﬁcation of

patterns in patients’actions, thoughts, feelings, experiences

and relationships; (d) an emphasis on past experiences; (e) a

focus on patients’interpersonal experiences; (f) an emphasis

on the therapeutic relationship; and (g) an exploration of

patients’wishes, dreams or fantasies. Their conclusions

were based in part on an expert-consensus study con-

ducted by Ablon and Jones (1998, 2005) who asked 11

leading psychoanalytic theoreticians and practitioners to

rate the 100 items of the Psychotherapy Process Q-Set

(PQS; Jones, 1985) with respect to how characteristic each

item was of the principles and activities ideally found in

psychodynamic/psychoanalytic therapies. The rank order

of items was very reliable (Cronbach’salpha=0.94),and

the most prototypical items comprised all interventions

mentioned above (except for the focus on patients’inter-

personal experiences), including several other features

such as a neutral attitude towards the patient and

discussing a patient’s sexual feelings and experiences

(Ablon & Jones, 1998, 2005).

Thus, there seems to be

sufﬁcient consensus regarding the distinctive features of

psychodynamic/psychoanalytic techniques, from both

empirical and conceptual points of view.

Starting from this consensual set of interventions, several

studies on short-term psychotherapies tested the effect of

psychodynamic or psychoanalytic techniques on outcomes,

mostly with positive results (Ablon & Jones, 1998; Ablon,

Levy, & Katzenstein, 2006; Gaston et al., 1998; Hilsenroth

et al., 2003; Owen & Hilsenroth, 2011; Slavin-Mulford,

Hilsenroth, Weinberger, & Gold, 2011). In an early study by

Ablon and Jones (1998), raters used the PQS to assess psycho-

therapy processes in STPP and CBT on the basis of verbatim

transcripts of selected therapy sessions. The item proﬁle of

each session was correlated with the proﬁle of prototypical

psychoanalytic techniques (see above), yielding an empirical

measure of the degree to which a session adhered to the

theoretical principles of psychoanalytic psychotherapy. In

one of two STPP samples, adherence to the psychoanalytic

prototype was positively correlated with several outcome

measures. Even more interesting, the application of psycho-

analytic techniques was also positively correlated with the

outcomeinCBTfordepression(Ablon&Jones,1998).Ina

subsequent study on STPP for panic disorder, Ablon et al.

(2006) showed that, although the psychotherapy that was de-

livered corresponded mainly to the cognitive–behavioural

prototype, only adherence to the psychoanalytic prototype

was positively correlated with self-reported outcome.

More recently, similar ﬁndings were reported by

Hilsenroth and colleagues (Hilsenroth et al., 2003; Owen

& Hilsenroth, 2011; Slavin-Mulford et al., 2011) using the

Comparative Psychotherapy Process Scale (Hilsenroth,

Blagys, Ackerman, Bonge, & Blais, 2005) to assess

psychodynamic–interpersonal and cognitive–behavioural

techniques from videotaped sessions. In their studies on

naturalistic STPP, they found that psychodynamic–

interpersonal techniques predicted positive outcomes

both in patients with major depression (Hilsenroth et al.,

2003) and in patients with an anxiety disorder (Slavin-

Mulford et al., 2011). Moreover, a study by Owen and

Hilsenroth (2011) revealed that patients with better

alliances beneﬁtted signiﬁcantly more from psychodynamic–

interpersonal interventions than patients with poorer

alliances, thereby replicating earlier ﬁndings by Gaston

et al. (1998). Taken together, there is growing evidence that

psychodynamic or psychoanalytic techniques are positively

associated with various outcome measures across different

treatment settings and disorders. Although the studies

reported above consistently focused on the immediate

effects of short-term treatments, it seems plausible to

hypothesize that psychoanalytic techniques are also

involved in facilitating sustained change in LTPP. Thus,

the effectiveness of intensive psychodynamic treatments

such as psychoanalytic therapy may not stem from the

higher number of sessions but from applying more psy-

choanalytic techniques (as compared with less intensive

or non-psychodynamic treatments).

The Current Study

The aim of the current study was to test these two rival

hypotheses empirically. Therefore, we employed data from

the MPS comparing psychoanalytic therapy (i.e., high-dose

LTPP) with psychodynamic therapy (i.e., low-dose LTPP)

and CBT in severely depressed patients. In previous publica-

tions using these data, Huber, Zimmermann, et al. (2012) and

Huber et al. (2013) found that patients receiving psychoana-

lytic therapy had more favourable symptom trajectories

across a 3-year follow-up period than patients receiving psy-

chodynamic therapy or CBT. In the present study, we tested

whether this differential long-term effect of psychoanalytic

In the prior literature, this PQS prototype has been labelled as both

‘psychodynamic’(Ablon & Jones, 1998) and ‘analytic’(Ablon &

Jones, 2005). In the following, we will use the term ‘psychoanalytic’

when referring to this speciﬁc set of techniques for two reasons: ﬁrst,

it contains several aspects that are less typical or even de-emphasized

in psychodynamic as compared with psychoanalytic treatments (e.g.,

neutral attitude towards the patient and focus on past experiences).

Second, psychodynamic treatments are more likely to include other

(e.g., supportive) techniques, suggesting that in psychoanalytic treat-

ments, this set of techniques might be applied in its most pure form.

This is in line with empirical results that have shown that psychoan-

alytic therapy sessions adhere more closely to the PQS prototype than

psychodynamic therapy sessions do (Ablon & Jones, 2005).

471Is It All about the Higher Dose?

therapy would be mediated by the number of sessions

and/or psychoanalytic technique. Our analyses were based

on three self-reported outcome measures that tracked

changes in depressive symptoms, interpersonal problems

and introject afﬁliation. These outcome measures were

selected to cover a broad spectrum of change processes,

including changes in primary symptoms as well as changes

in more enduring interpersonal and intrapsychic aspects of

patients’personalities. Because the previous literature has

provided support for both dose-based and technique-based

explanations for the effectiveness of LTPP, we had no apriori

assumptions about which of the two hypotheses would be

more valid. Finally, note that the main focus of this study

was on the indirect effects of dose and psychoanalytic

technique on the long-term outcomes across the 3-year

follow-up period. This is because the immediate treatment

outcome might be confounded with the mediating variables

(e.g., some treatments might have been shorter because

symptoms improved earlier), whereas the long-term out-

come cannot (i.e., after a treatment is completed, the number

of sessions and the employed techniques are ﬁxed and

cannot be inﬂuenced by future changes in symptoms). How-

ever, for the sake of completeness, we also tested respective

indirect effects on the immediate post-treatment outcome.

METHOD

Study Design

The MPS is a comparative quasi-experimental study of

psychoanalytic therapy, psychodynamic therapy and

CBT. It was designed to maximize external validity by

examining non-manualized and representative psycho-

therapies under the conditions of day-to-day practice con-

ducted by experienced psychotherapists, while improving

internal validity by recruiting a diagnostically homoge-

neous sample, blinding interviewers and raters with

regard to treatment modality and randomly allocating

participants to certain treatment groups. That is, the

MPS shares some characteristics with effectiveness studies

(i.e., pragmatic studies that investigate treatments under

‘real-world’conditions) and some characteristics with

efﬁcacy studies (i.e., explanatory studies that investigate

treatments under ideal conditions; Thorpe et al., 2009).

The term ‘quasi-experimental’denotes the fact that,

due to limited ﬁnancial resources, the randomized allo-

cation began with psychoanalytic and psychodynamic

therapies and included CBT later on. Speciﬁcally, CBT

was included by suspending the random assignment

to the psychoanalytic versus psychodynamic therapies

and assigning patients directly to CBT until the CBT

condition contained as many patients as the other two

conditions. That is, participants had equal chances of

being assigned to psychoanalytic versus psychody-

namic therapy throughout the study (i.e., they were

fully randomly allocated to these two treatment condi-

tions), whereas the chance of being assigned to CBT

was different (at least in the earlier phase of the

study).

The study was conducted at the Department

of Psychosomatic Medicine and Psychotherapy,

Technische Universität München (TUM, Germany).

The study protocol was approved by the Ethics Com-

mittee of the TUM. More detailed descriptions of the

studydesigncanbefoundintheworksofHuber,

Henrich, et al. (2012), Huber, Zimmermann, et al.

(2012) and Huber et al. (2013).

Sample

The full intent-to-treat sample consisted of 100 patients,

with 35 patients receiving psychoanalytic therapy, 31

patients receiving psychodynamic therapy and 34 patients

receiving CBT. Inclusion criteria required participants to

have the following: (a) a Beck Depression Inventory

(BDI) total score of at least 16 (which was established as a

cut-off for screening purposes) and (b) a primary diagnosis

of a major depressive disorder with a current moderate or

severe episode (International Classiﬁcation of Diseases, 10th

revision, diagnosis F 32.1/2 or F 33.1/2 or Diagnostic and

Statistical Manual of Mental Disorders, fourth edition, diag-

nosis 296.22/23 or 296.32/33) or a double depression char-

acterized by both dysthymic disorder and a current

moderate or severe depressive episode. Exclusion criteria

were bipolar affective disorder, depression due to somatic

illnesses or diseases of the brain, alcohol or substance

dependence, psychotherapy during the past 2 years and

concurrent antidepressants. For a more detailed descrip-

tion of the intent-to-treat sample, see Huber, Henrich,

et al. (2012) and Huber, Zimmermann, et al. (2012).

This is essentially a trade-off as some features (e.g., the lack of treat-

ment manuals) may represent a strength in terms of generalizability

to “real-world”conditions (e.g., therapists usually do not adhere to

manuals) but may represent a limitation in terms of the interpretabil-

ity of the results (e.g., it might be unclear what the therapists actually

did in treatment). In this regard, one should keep in mind that, at the

time the MPS started, manuals for LTPP were not available. One

might even argue that the usual format of manuals, precisely describ-

ing therapeutic action in a determined order session by session, is at

odds with the psychoanalytic approach to psychotherapy and impos-

sible to realize for long-term treatments.

Note that at the beginning of the study, an external randomization

board was established to examine whether a patient could be ran-

domly allocated to the experimental groups (taking into account

criteria such as strong preference for a speciﬁc therapy and time re-

sources). Participants were included only if they met this criterion.

After the inclusion of CBT, the external randomization board con-

ﬁrmed that all participants who had been randomized prior to the in-

clusion of CBT would also have been eligible for CBT.

472 J. Zimmermann et al.

All sessions of all treatments were supposed to be

audio-recorded by therapists. The three sessions in the

middle of each treatment were selected to be analysed in

terms of psychotherapy processes (see below). In the

following analyses, we included only the following

patients: (a) those who provided ratings on outcome

measures at post-treatment or follow-up and (b) those

whose middle sessions were audio-recorded with sufﬁ-

cient quality. Four patients dropped out during treatment

and did not provide further ratings on outcome measures.

In 19 cases, audio recordings of middle sessions were not

available or had to be excluded due to low audio quality.

Thus, the ﬁnal sample consisted of 77 patients, with 27

patients receiving psychoanalytic therapy, 26 patients

receiving psychodynamic therapy, and 24 patients

receiving CBT. As shown in Table 1, there were no

signiﬁcant differences in demographic or diagnostic

variables at pre-treatment between groups except for

age. Therefore, we included age as a covariate in the

following analyses.

Treatments

Treatments were ‘real-world’, non-manualized and

representative psychotherapies provided in the German

healthcare system by 21 highly experienced psychothera-

pists (the mean duration of their psychotherapeutic

practice was 15 years; their mean age was 47 years); 14

therapists delivered both psychoanalytic and psycho-

dynamic therapies, and seven therapists delivered CBT.

As we had to exclude 23 patients from the present

analyses (see above), the number of therapists involved

in this study was reduced to 19. There were no signiﬁcant

differences in training, expertise or experience between

the therapists who administered the different treatment

modalities.

Psychoanalytic therapy is deﬁned as an interpretative,

insight-oriented approach that is aimed at modifying

maladaptive representations of the self and others that

lie at the root of psychopathology (Fonagy & Kächele,

2009). It involves careful attention to the therapist–patient

interaction with thoughtfully timed interpretation of

transference and resistance (Gabbard, 2004). According

to the German Psychotherapy Guidelines (Rüger, Dahm,

& Kallinke, 2003), the average dose is between 160 and

240 sessions; session frequency is two to three sessions

per week with the patient lying on a couch. In the present

study, the dose ranged from 58 to 356 sessions (M= 241.3,

SD = 89.9) and the duration from 12 to 72 months

(M= 39.3, SD = 16.6).

Psychodynamic therapy is based on the same principles of

theory and technique but is more limited in the depth of

the therapeutic process and in its goals by focusing on

symptom-sustaining here-and-now conﬂicts without

enhancing regression in the therapeutic process. Its mean

dose is between 50 and 80 sessions, and the session fre-

quency is one session per week with the patient sitting

in a face-to-face position (Rüger et al., 2003). In the present

study, the dose ranged from 18 to 218 sessions (M= 85.4,

SD = 56.5) and the duration from 11 to 118 months

(M= 32.6, SD = 24.2).

Cognitive–behavioural therapy comprises therapeutic

modalities developed on the basis of a psychology of

learning and social psychology and combines cognitive

and behavioural techniques to modify distorted or

maladaptive cognitions and facilitate positive thoughts

and behaviours (Hollon & Beck, 2004). The therapist

employs a goal-directed, problem-focused and directive

approach, often using homework assignments to stabi-

lize therapeutic progress. According to the German

Psychotherapy Guidelines, the average dose is between

45 and 60 sessions, and the session frequency is one

session per week (Rüger et al., 2003). In the present

study, the dose ranged from 19 to 100 sessions

Table 1. Pre-treatment sample description

Psychoanalytic therapy Psychodynamic therapy Cognitive–behavioural therapy

Sample size 27 26 24

Age (years) 31.0 (6.0) 35.7 (7.3) 34.3 (5.5) F(2, 74) = 3.83*

Female 19 16 21 χ

(2) = 4.36

Relationship status χ

(4) = 2.82

Single 12 10 7

Separated 10 7 9

Partnership 5 9 8

Duration of illness in months 73.5 (85.6) 67.5 (87.5) 61.4 (74.8) F(2, 74) = 0.13

Severe depressive episode 12 9 9 χ

(2) = 0.57

Double depression 17 13 10 χ

(2) = 2.37

Personality disorder 9 8 7 χ

(2) = 0.11

*p<0.05.

473Is It All about the Higher Dose?

(M= 50.8, SD = 23.3) and the duration from 8 to

78 months (M= 29.0, SD =15.5).

Previous publications on the full intent-to-treat sample

have shown that the integrity of the three treatments

was adequate (Huber, Zimmermann, et al., 2012; Huber

et al., 2013). Speciﬁcally, the therapists who delivered

the psychoanalytic and psychodynamic treatments rated

each treatment every 6 months on various parameters

(cf. Beenen & Stoker, 1996), and results showed that in

the psychoanalytic treatments, the session frequency

was higher, patients were lying down rather than

sitting, transference to the therapist was stronger and

the employed techniques were insight oriented rather

than supportive (Huber et al., 2013). Moreover, 50% of

the treatments from all three experimental groups were

selected at random, and one audiotaped session from

the middle of each treatment was assessed by trained

raters using the PQS (see below). Results showed that

psychoanalytic techniques were more salient in psycho-

analytic therapy sessions as compared with the other

two conditions, and cognitive–behavioural techniques

were more salient in CBT sessions as compared with

the other two conditions (Huber, Zimmermann, et al.,

2012). The present study extended these ﬁndings on

treatment integrity by analysing a larger sample of

middle sessions.

Measures

Depressive Symptoms

We assessed the severity of depressive symptoms using

the BDI (Beck et al., 1961; German version: Hautzinger,

Bailer, Worall, & Keller, 1995). The BDI contains 21

multiple-choice items covering a broad spectrum of

depressive symptoms. Respondents are asked to indicate

the intensity of each symptom during the last week. The

BDI is one of the most common self-report measures of

depression, generally showing good reliability and

validity (Beck, Steer, & Garbin, 1988).

Interpersonal Problems

The severity of interpersonal problems was assessed by

means of the Inventory of Interpersonal Problems—

Circumplex (IIP-C; Alden, Wiggins, & Pincus, 1990;

German version: Horowitz, Strauß, & Kordy, 2000). The

IIP-C contains 64 items describing deﬁcits or excesses

related to speciﬁc interpersonal behaviours. For each item,

respondents are asked to indicate the amount of distress

on a 5-point scale ranging from 0 (not at all)to4

(extremely). We used the global score, with high values

indicating more interpersonal distress (Tracey, Rounds,

& Gurtman, 1996). The IIP-C is a widely used measure

of interpersonal problems with good psychometric

properties and is well accepted across theoretical and

therapeutic orientations (Hughes & Barkham, 2005).

Introject Afﬁliation

Introjects (i.e., the habitual way people treat themselves)

were assessed by means of the introject surface of the

INTREX short form (Benjamin, 1983; German version:

Tress, 1993). The INTREX is a self-report measure based

on the Structural Analysis of Social Behavior (SASB;

Benjamin, 1974). The introject surface of the SASB cluster

model combines the dimensions of afﬁliation (active self-

love vs self-attack) and interdependence (self-emancipation

vs self-control) into eight clusters. Participants are asked to

rate how they treat themselves at their best and at their

worst. Each of the eight SASB clusters is measured with a

single item for best and worst, respectively, using a scale

from 0 to 100. For instance, the positive pole of the afﬁliation

dimension (i.e., active self-love) is represented by the item

‘I tenderly, lovingly cherish myself’,andthenegative

pole (i.e., self-attack) is represented by the item ‘Without

considering what might happen, I hatefully reject and de-

stroy myself’. Individual items can be aggregated into

‘vector scores’for measuring individual differences in

introject afﬁliation and autonomy (Pincus, Newes,

Dickinson, & Ruiz, 1998). In the following analyses, we

used only the vector score for introject afﬁliation at worst.

Previous research has shown that this dimension is at the

core of psychopathology (Monsen et al., 2007; Pincus,

Gurtman, & Ruiz, 1998) and can be considered to be a

primary target of psychotherapeutic change (e.g., Bedics,

Atkins, Comtois, & Linehan, 2012).

Psychoanalytic Technique

Psychoanalytic technique was assessed with the PQS

(Jones, 1985; German version: Albani et al., 2000). The

PQS is a pantheoretically developed instrument that

provides a comprehensive language and rating system

for describing psychotherapy processes. It consists of 100

items (i.e., the ‘Q set’) covering actions, behaviours and

thoughts of both therapist and patient in individual and

dyadic terms (e.g., ‘Therapist suggests that patient accept

responsibility for his/her problems’,‘Patient expresses

angry or aggressive feelings’or ‘Sexual feelings and

experiences are discussed’). Raters listen to audio tapes

(or read verbatim transcripts) of a full therapy session

and are asked to sort the 100 items in the Q set on a con-

tinuum from least characteristic (Category 1) to most

characteristic (Category 9). The middle pile (Category 5)

is used for items deemed either neutral or irrelevant to

the particular hour being rated. Raters are forced to

adhere to a ﬁxed normal distribution, thereby ensuring

multiple evaluations of items and attenuating rater biases

such as response sets and halo effects. Afterwards, the

proﬁle of a given session can be correlated with proﬁles

474 J. Zimmermann et al.

of ‘ideal’psychotherapeutic processes that have been

developed on the basis of expert ratings for a range of

therapeutic approaches (Ablon & Jones, 1998, 2002).

These ‘prototypes’can be conceived of as the shared

knowledge of leading experts, delineating which tech-

niques are consensually regarded as characteristic versus

uncharacteristic of a speciﬁc approach (see above). The

(Fisher-ztransformed) correlations of the proﬁle with

these prototypes represent the extent to which a speciﬁc

therapeutic approach is implemented in a given session

and can be conceived of as a measure of adherence.

We computed correlations of the proﬁles with the psy-

choanalytic prototype (PA prototype). Across samples

and therapies, PA prototype mean scores usually range

from 0 to 0.50 (Ablon & Jones, 1998, 2005), with higher

values reﬂecting techniques that are more typical of a

psychoanalytic approach to psychotherapy. Note that

we did not include correlations of the proﬁles with the

cognitive–behavioural prototype in the main analyses

as we had no a priori hypotheses about the effect of

cognitive–behavioural techniques on outcomes within

long-term treatments. However, for the sake of com-

pleteness, we will report the respective results in a foot-

note. The PQS has demonstrated reliability and validity

across a variety of different treatments including psy-

choanalytic, psychodynamic, cognitive–behavioural and

other therapies (Ablon & Jones, 2002, 2005; Jones,

Cumming, & Horowitz, 1988; Jones & Pulos, 1993).

In the present study, PQS ratings were based on

audio recordings of three therapy sessions from the

middle of each treatment. We favoured middle sessions

over early sessions because it seemed unreasonable to

assume that psychoanalytic techniques such as focusing

on sexual experiences, making transference interpreta-

tions or reconstructing childhood experiences would

ever be employed during the early sessions of a long-

term treatment. In fact, using such techniques during

the early sessions might even be considered to be coun-

terproductive (e.g., Gabbard & Horowitz, 2009; Krause,

2009). More generally, we argue that when assessing

speciﬁc therapeutic processes (as compared with unspe-

ciﬁc processes such as therapeutic alliance), middle

sessions might be more representative of the intense

working phase of a therapy as compared with early

sessions (which are usually about deﬁning roles, estab-

lishing an alliance etc.) or late sessions (which are

usually about reﬂecting on achievements, the upcoming

termination of therapy or expectations for the future).

The main drawback of using middle sessions is that

the therapeutic processes are probably inﬂuenced by

prior symptom change. The speciﬁc sessions that were

analysed were selected purely on arithmetic grounds

(e.g., in a treatment with a total number of 100 sessions,

we selected the 49th, 50th and 51st sessions). Indepen-

dent ratings of three sessions were available in 73 cases.

In three cases, ratings were available from two sessions,

and in one case, ratings were available from one session.

Each of the 226 sessions was rated by one of two

raters, both blind to any information about patient,

therapist, treatment and outcome.

The sessions were

given to the raters in a random order to avoid memory

effects. Inter-rater reliability was assessed in a subsample

of 29 sessions. Proﬁle correlations ranged from 0.37 to

0.87 with a median value of 0.65. The intraclass correla-

tion coefﬁcient (ICC) for the PA prototype score was

ICC(2, 1) = 0.59, p<0.001, which is ‘fair’according to

Cicchetti’s (1994) guidelines and typical for psychody-

namic adherence measures (e.g., Leichsenring, Salzer,

et al., 2013). To further enhance reliability, we combined

PA prototype scores from the three sessions per treat-

ment. As sessions from the same treatment were

assessed by different raters (due to the randomized pre-

sentation of sessions), this kind of data aggregation was

chosen to help cancel out rater-speciﬁc measurement

error. The internal consistency of the average PA proto-

type score was α= 0.78. Thus, treatment differences in

psychoanalytic technique were relatively stable across

the three middle sessions. All subsequent analyses were

based on the average PA prototype score.

Statistical Analyses

We used a mediation analysis to test whether the differen-

tial long-term effect of psychoanalytic therapy was due to

the number of sessions and/or the psychoanalytic tech-

nique (Hayes, 2009; MacKinnon, 2008). In its simplest

form, a mediation analysis requires two linear regression

models to be computed to assess the effect of the indepen-

dent variable on the mediating variable (Path a) and the

effect of the mediating variable on the outcome variable

while controlling for the independent variable (Path b).

After that, the signiﬁcance of the indirect effect can be

directly tested by comparing the product of the two

regression coefﬁcients (ab) with its standard error. Because

the sampling distribution of the product deviates from a

normal distribution, it is recommended that bootstrap

(Preacher & Hayes, 2004) or Monte Carlo (Preacher &

Selig, 2012) conﬁdence limits be computed for signiﬁcance

testing (MacKinnon, Lockwood, & Williams, 2004).

We expanded the simple mediation analysis in several

respects (Figure 1): ﬁrst, because the independent variable

Prior to rating these sessions, raters were trained by a psychiatrist

and psychoanalyst (the second author) who was originally certiﬁed

as a PQS rater by Enrico E. Jones in 2002. Raters had no psychiatric

or psychotherapeutic background. During their training, they had

to rate ﬁlm sequences that included problematic dyadic situations

(e.g., speciﬁc sequences from the ﬁlm ‘The Silence of the Lambs’,

1991).

475Is It All about the Higher Dose?

was multicategorical, we used dummy coding to deﬁne

psychodynamic therapy and CBT as dummy variables

and psychoanalytic therapy as the reference group (Hayes

& Preacher, 2011). Thus, we will present two separate

mediation analyses, each time estimating the relative

indirect effect of psychoanalytic therapy as compared

with a different ‘control’condition (i.e., psychodynamic

therapy or CBT). Second, because our hypothesis involved

two rival intervening processes, we used a multiple-

mediator model (Preacher & Hayes, 2008). This requires

that two separate regression analyses be computed to

assess the effects of the dummy variables on both the

number of sessions and the psychoanalytic technique.

Moreover, we included age and the pre-treatment scores

of the outcome variable as covariates (not shown in

Figure 1).

This resulted in the following regression

equations for Path a:

DOSEi¼β00 þβ01*AGEiþβ02 *PRE-OUTCOMEi

þβ03*CBTiþβ04 *PDTiþεi

PAPROTOTYPEi¼β10 þβ11 *AGEiþβ12*PRE-OUTCOMEi

þβ13*CBTiþβ14 *PDTiþεi;

where DOSE

and PA PROTOTYPE

represent the number of

sessions and the salience of the psychoanalytic technique in

sessions of patient i’s treatment (i.e., the mediating variables),

CBT and PDT are the dummy-coded independent variables,

AGE and PRE-OUTCOME are covariates and β

,β

and β

are the regression coefﬁcients of interest, representing

the relative effects of psychoanalytic therapy on the mediat-

ing variables (Figure 1). These models were ﬁt with R (R Core

Team, 2013) using ordinary least square estimation and

robust standard errors (Long & Ervin, 2000).

Third, as we were interested in symptom change after

treatment, we used a multilevel mediator model (Krull &

MacKinnon, 2001). Because only the outcome variable

varied within persons, the appropriate model type was

upper-level mediation (2 →2→1). Upper-level mediation

can be implemented by ﬁtting a multilevel model with

years after treatment as the Level-1 predictor and dummy

variables for the treatment groups as well as the mediating

variables as the Level-2 predictors. To control for indi-

vidual differences in age and pre-treatment scores on the

outcome variables, we included the respective variables

as covariates at Level 2. Moreover, we included random

effects of the intercept and the slope to allow for individual

differences in the trajectories of change.

This resulted in

the following multilevel regression model for Path b:

We imputed three missing data points in pre-treatment INTREX

scores using the expectation–maximization algorithm in PASW 18

(IBM Corporation, Armonk, NY; Schafer & Graham, 2002). Data im-

putation was based on the available demographic and clinical pre-

treatment data (Table 1) as well as the pre-treatment BDI and IIP-C

scores.

We acknowledge that outcome differences between patients (as well

as differences in dose and technique) may be due to differences be-

tween therapists and that ignoring this nested factor may bias param-

eter estimates (Wampold & Serlin, 2000). Thus, we repeated all

analyses with the 19 therapists included as random effects at Level

2 (in Path aregressions) or as random effects of the intercept and

slope at Level 3 (in Path bregressions). The results were virtually

the same, and thus, we decided to present the simpler model.

Figure 1. Upper-level mediation with a multicategorical independent variable and two mediators at Level 2

(1)

476 J. Zimmermann et al.

The ﬁrst row includes the Level-1 submodel, where

OUTCOME

represents the value of the outcome variable

of patient iat time j, TIME

represents the number of years

after the end of treatment at that time and the parame-

ters π

and π

represent patient i’s estimated post-

treatment status (i.e., Level-1 intercept) and estimated

rate of change per year during the follow-up period (i.

e., Level-1 slope), respectively. The subsequent rows

present the Level-2 submodel for predicting interindi-

vidual differences in post-treatment status and rate of

change during follow-up. Speciﬁcally, γ

and γ

repre-

sent the estimated post-treatment status and rate of

change of patients receiving psychoanalytic therapy; γ

to γ

and γ

to γ

represent the unique inﬂuences of

the covariates, treatment groups and mediators on the

Level-1 parameters; and ζ

and ζ

represent those

portions of the Level-1 parameters that remain unex-

plained at Level 2 (i.e., the random effects). In line with

our hypothesis, we mainly focused on γ

and γ

representing the speciﬁc effects of dose and psychoanalytic

technique on the slope of the outcome (i.e., on its rate of

change per year after treatment). However, we also con-

sidered γ

and γ

, representing the speciﬁc effects of the

mediators on the intercept of the outcome (i.e., on its

expected value at post-treatment; Figure 1). All models

were ﬁt using the software package HLM 6.08 (Scientiﬁc

Software International, Inc., Skokie, IL) with restricted max-

imum likelihood estimation and robust standard errors.

Finally, we computed the products of the regression

coefﬁcients from the single-level models (Path a) and the

multilevel model (Path b) and assessed their statistical

signiﬁcance using Monte Carlo conﬁdence intervals (CI;

Preacher & Selig, 2012; Selig & Preacher, 2008). All media-

tion analyses were computed separately for each of the

three outcome measures.

RESULTS

Descriptive Analyses

Table2summarizesthedescriptive characteristics and

intercorrelations of the mediators and outcome vari-

ables in the combined sample. Unsurprisingly, the two

mediators were moderately correlated, which means

that patients who received more sessions were likely

to receive more psychoanalytic interventions. Note,

however, that dose and psychoanalytic technique were

uncorrelated within treatment conditions (i.e., when

controlling for the two dummy-coded treatment

variables; r=0.03, p= 0.81; not shown in Table 2). Both

mediators were generally not associated with the

pre-treatment scores of the outcome measures but

tended to be negatively associated with the outcome

measures at several follow-up occasions. Taken

together, these ﬁndingssuggestthatbothvariablesare

indeed plausible candidates for explaining variance in

long-term outcomes.

Total Effects of Psychoanalytic Therapy on Outcome

In a preliminary step, we looked for the differential total ef-

fects of psychoanalytic therapy on the intercept and slope in

the multilevel analyses. Therefore, we ﬁtmodelsinwhich

the mediating variables were omitted from Equation 2.

The results are presented in Table 4 in the ﬁrst column for

each outcome measure, respectively. The expected mean

scores of outcome measures directly after treatment for pa-

tients who received psychoanalytic therapy were clearly in

the non-clinical range (γ

). Moreover, these patients

showed further improvements during the 3-year follow-

up period as their depressive symptoms and interpersonal

problems decreased, and their habitual way of treating

themselves became more afﬁliative (γ

). When considering

the total effects of psychoanalytic therapy as compared with

CBT, the results suggest that participants receiving CBT had

more interpersonal problems and more hostile introjects

directly after treatment (γ

) and tended to show a less

advantageous trajectory of change in terms of depressive

symptoms and interpersonal problems during the 3-year

follow-up period (γ

). A similar picture emerged for the

total effects of psychoanalytic therapy as compared with

psychodynamic therapy: participants receiving psychody-

namic therapy had more hostile introjects directly after

treatment (γ

) and were less likely to improve during

follow-up in terms of depressive symptoms (γ

). However,

psychoanalytic and psychodynamic therapies did not differ

with regard to changes in interpersonal problems. In sum,

these ﬁndings are in line with the differential effectiveness

of psychoanalytic therapy that has been reported for the full

intent-to-treat sample of the MPS (for more detailed results

on effect sizes and response rates, see Huber, Zimmermann,

et al., 2012; Huber et al., 2013).

OUTCOMEij ¼π0iþπ1i*TIMEij þεij

π0i¼γ00 þγ01*PRE-OUTCOMEiþγ02 *AGEiþγ03*CBTiþγ04*PDTiþγ05 *DOSEiþγ06*PA PROTOTYPEiþζ0i

π1i¼γ10 þγ11*PRE-OUTCOMEiþγ12 *AGEiþγ13*CBTiþγ14*PDTiþγ15 *DOSEiþγ16*PA PROTOTYPEiþζ1i:

(2)

477Is It All about the Higher Dose?

Explaining Outcome Differences between Psychoanalytic

Therapy and Cognitive–Behavioural Therapy

Table 3 summarizes the results of the regression anal-

yses for Path a. When focusing on the relative differ-

ences between psychoanalytic therapy and CBT, the

results show that psychoanalytic therapy comprised a

strikingly greater number of sessions than CBT (β

and psychoanalytic interventions were far more typi-

cal for psychoanalytic therapy than for CBT (β

). This

was true irrespective of which pre-treatment outcome

variable was controlled for. However, it should be

noted that, on average, psychoanalytic techniques

were less salient in psychoanalytic sessions than in

former studies, as B=0.26 for the intercept β

clearly lower than the average PA prototype score

for the psychoanalytic therapies reported by Ablon

and Jones (2005).

The results for Path bare presented in Table 4 in the

second column of each respective outcome measure.

When focusing on the long-term outcome during follow-

up, we found that when controlling for age, pre-treatment

status, treatment group and psychoanalytic technique, the

number of sessions failed to reach statistical signiﬁcance

in predicting the change in outcome measures (γ

). By

contrast, psychoanalytic technique was a unique predictor

of follow-up changes in depressive symptoms and intro-

ject afﬁliation (γ

); this means that patients who received

a larger number of psychoanalytic interventions had

more favourable symptom trajectories and developed

amoreafﬁliative introject during follow-up, even when

controlling for age, pre-treatment status, treatment

group and number of sessions.

Note that the direct

To gauge the practical implications of these effects, we computed the

change in BDI and INTREX scores during the 3-year follow-up period

that would be expected from a difference of 1 SD in the PA prototype

score. Everything else being equal, patients who received more psy-

choanalytic techniques (+0.13) were expected to show a reduction of

1.63 BDI points and an increase of 15.2 INTREX points at the 3-year

follow-up.

Table 2. Descriptive statistics and intercorrelations of target variables

Descriptive statistics Correlations with mediators

NRange MSD Dose Technique

Mediators

Dose 77 0.18 —3.56 1.29 1.05

Technique 77 0.18 —0.48 0.19 0.13 0.33**

BDI

Pre 77 16 —47 25.43 8.21 0.05 0.09

Post 75 0 —28 6.93 7.00 0.03 0.08

Fu 1 72 0 —41 8.04 8.14 0.09 0.18

Fu 2 63 0 —43 8.41 9.25 0.21

0.30*

Fu 3 66 0 —36 7.58 7.98 0.31* 0.28*

IIP-C

Pre 77 0.33 —2.63 1.75 0.43 0.02 0.06

Post 75 0.08 —2.41 1.27 0.57 0.26* 0.22

Fu 1 70 0.13 —2.77 1.31 0.62 0.40*** 0.22

Fu 2 61 0.11 —2.94 1.33 0.57 0.40*** 0.30*

Fu 3 64 0.20 —2.43 1.14 0.59 0.42*** 0.25*

INTREX

Pre

†

77 168 —145 55.2 59.8 0.05 0.06

Post 77 140 —168 4.9 79.7 0.26* 0.06

Fu 1 73 168 —168 9.4 71.4 0.20

0.01

Fu 2 60 105 —150 11.6 71.0 0.29* 0.25

Fu 3 57 115 —164 30.4 73.3 0.29** 0.19

Dose = number of se ssions/100. Technique= average Fisher-ztransformed proﬁle correlations with the psychoanalytic Psychotherapy Process

Q-Set prototype. BDI = Beck Depression Inventory. IIP-C = Inventory of Interpersonal Problems—Circumplex. INTREX = introject afﬁliation at

worst. Pre = pre-treatment score. Post = post-treatment score. Fu 1 = 1 year after termination. Fu 2 = 2 years after termination. Fu 3= 3 year s

after termination.

†

Three missing values were imputed on the basis of pre-treatment data using the expectation–maximization algorithm in PASW 18 (IBM Corporation).

p<0.10,

*p<0.05,

**p<0.01,

***p<0.001.

478 J. Zimmermann et al.

effects of psychoanalytic therapy on long-term outcome

were generally not signiﬁcant (γ

), suggesting that the

mediating variables were successful in explaining the

total effect.

Although not central to this paper, we also consi-

dered effects of the mediating variables on the outcome

measures directly after treatment. In this regard, we

found that a higher dose was uniquely associated with

fewer interpersonal problems and more afﬁliative

introjects directly after treatment (γ

), whereas psycho-

analytic technique was unrelated to post-treatment

outcome (γ

). This suggests that, irrespective of age,

pre-treatment status, treatment group and applied

techniques, attending a larger number of sessions helped

patients reduce their interpersonal distress and their

hostility towards themselves during treatment. Again,

the direct effects of psychoanalytic therapy were not sig-

niﬁcant (γ

Finally, we tested the mediation hypotheses using

95% Monte Carlo CIs for the products of the coefﬁ-

cients of Paths aand b. As shown in the third column

of Table 5, only psychoanalytic technique but not

number of sessions mediated the differential long-term

effect of psychoanalytic therapy on depressive symp-

toms. For interpersonal problems, neither psychoana-

lytic technique nor number of sessions turned out to

be a signiﬁcant mediator. As we did not ﬁnd a differen-

tial long-term effect of psychoanalytic therapy for intro-

ject afﬁliation, testing for mediation might seem

unreasonable because there was no total effect that

could be explained. However, Hayes (2009) noted that

even in such cases, mediation can be present as other

mediators might be operating in the opposite direction,

thereby cancelling each other out and reducing the total

effect to zero. Indeed, we found that psychoanalytic

therapy had an indirect effect on long-term outcome

through psychoanalytic technique but not through

number of sessions. Thus, our data were at least par-

tially in line with the hypothesis that the sustained

change induced by psychoanalytic therapy was due to

psychoanalytic technique and not due to a higher num-

ber of sessions. However, it should be noted that we

found speciﬁc indirect effects of psychoanalytic therapy

on post-treatment interpersonal problems and post-

treatment introject transmitted by dose (not shown in

Table 5). Thus, the advantage of psychoanalytic therapy

Table 3. Predicting dose and psychoanalytic technique (Path a)

Dose Psychoanalytic technique

BSEβBSEβ

BDI as covariate

Intercept

†

2.430*** 0.176 β

0.259*** 0.020

Pre-treatment score

‡

0.016

0.008 0.126 β

0.002 0.002 0.099

Age

‡

0 0.010 0.001 β

0.003 0.002 0.155

CBT (vs PAT) β

1.931*** 0.186 0.857 β

0.149*** 0.031 0.543

PDT (vs PAT) β

1.588*** 0.207 0.719 β

0.072* 0.035 0.267

IIP-C as covariate

Intercept

†

2.424*** 0.179 β

0.258*** 0.020

Pre-treatment score

‡

0.029 0.162 0.012 β

0.004 0.039 0.014

Age

‡

0.004 0.010 0.027 β

0.003 0.002 0.177

CBT (vs PAT) β

1.923*** 0.188 0.853 β

0.148*** 0.031 0.538

PDT (vs PAT) β

1.576*** 0.217 0.714 β

0.072* 0.034 0.266

INTREX as covariate

Intercept

†

2.434*** 0.178 β

0.257*** 0.021

Pre-treatment score

‡

0.001 0.001 0.056 β

000.031

Age

‡

0.004 0.010 0.023 β

0.003 0.002 0.178

CBT (vs PAT) β

1.933*** 0.186 0.858 β

0.147*** 0.031 0.536

PDT (vs PAT) β

1.596*** 0.210 0.723 β

0.070

0.037 0.260

N= 77. Dose = number of sessions/100. Psychoanalytic techniqu e = average Fisher-ztransformed proﬁle correlations with the psychoanalytic Psycho-

therapy Process Q-Set prototype. CBT = cognitive–behavioural therapy. PAT = psychoanalytic therapy. PDT = psychodynamic therapy. BDI = Beck De-

pression Inventory. IIP-C = Inventory of Interpersonal Problems—Circumplex. INTREX = introject afﬁliation at worst.

†

Represents the estimate for patients receiving psychoanalytic therapy.

‡

Mean centred.

p<0.10,

*p<0.05,

**p<0.01,

***p<0.001.

479Is It All about the Higher Dose?

Table 4. Predicting outcome and stability of outcome (Path b)

BDI IIP-C INTREX

Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

Fixed effects, intercept

Mean post-treatment status

†

6.781*** (1.117) 5.610*** (1.375) 1.120*** (0.092) 1.345*** (0.124) 29.823* (13.735) 3.273 (15.271)

Pre-treatment score

‡

0.327* (0.132) 0.346* (0.131) 0.803*** (0.101) 0.798*** (0.097) 0.638*** (0.106) 0.610*** (0.098)

Age

‡

0.043 (0.115) 0.040 (0.122) 0.006 (0.008) 0.003 (0.008) 3.337** (0.988) 3.385** (0.959)

CBT (vs PAT) γ

1.842 (1.937) 3.751 (2.583) 0.403** (0.124) 0.008 (0.198) 42.577* (17.597) 1.256 (25.444)

PDT (vs PAT) γ

0.079 (1.676) 1.540 (2.136) 0.166 (0.126) 0.129 (0.177) 37.819* (18.264) 0.445 (22.847)

Dose

‡

1.017 (0.988) 0.161* (0.076) 25.763* (10.587)

Psychoanalytic technique

‡

0.271 (6.866) 0.611 (0.446) 43.286 (50.279)

Fixed effects, slope

Mean rate of change

†

0.667* (0.313) 0.361 (0.481) 0.059** (0.021) 0.040 (0.025) 6.616* (3.181) 8.140 (4.954)

Pre-treatment score

‡

0.039 (0.053) 0.058 (0.052) 0.009 (0.033) 0.009 (0.034) 0.073* (0.035) 0.067* (0.033)

Age

‡

0.023 (0.045) 0.033 (0.045) 0.001 (0.002) 0.001 (0.003) 0.500 (0.359) 0.610

(0.354)

CBT (vs PAT) γ

1.576

(0.935) 0.209 (1.114) 0.057* (0.028) 0.027 (0.042) 6.839 (5.258) 8.491 (8.611)

PDT (vs PAT) γ

1.321* (0.554) 0.039 (0.722) 0.039 (0.043) 0.013 (0.045) 0.876 (5.819) 2.350 (6.854)

Dose

‡

0.595 (0.361) 0.016 (0.019) 3.371 (2.868)

Psychoanalytic technique

‡

4.184** (1.554) 0.043 (0.132) 38.895* (17.049)

Models with BDI were based on 276 measurement points, models with IIP-C on 270 and models with INTREX on 267. Fixed effects are unstandardized regression coefﬁcients with the

respective robust standard errors in brackets. All models were ﬁt with HLM 6.08 (Scientiﬁc Software International, Inc.) using restricted maximum likelihood estimation. Model 1 = con-

ditional model with treatment effects and covariates. Model 2 = ﬁnal conditional model with mediators. Dose = number of sessions/100. Psychoanalytic technique = average Fisher-z

transformed proﬁle correlations with the psychoanalytic Psychotherapy Process Q-Set prototype. CBT = cognitive–behavioural therapy. PAT = psychoanalytic therapy. PDT = psychody-

namic therapy. BDI = Beck Depression Inventory. IIP-C = Inventory of Interpersonal Problems—Circumplex. INTREX = introject afﬁliation at worst.

†

Represents the estimate for patients receiving psychoanalytic therapy.

‡

Grand mean centred.

p<0.10,

*p<0.05,

**p<0.01,

***p<0.001.

480 J. Zimmermann et al.

Table 5. Relative speciﬁc indirect effects transmitting the effect of psychoanalytic therapy to long-term outcome

Psychoanalytic therapy versus CBT Psychoanalytic versus psychodynamic therapy

Path a(SE) Path b(SE)ab [95% CI] Path a(SE) Path b(SE)ab [95% CI]

BDI as outcome

PAT →dose →slope 1.931 (0.186) 0.595 (0.361) 1.149 [2.572, 0.216] 1.588 (0.207) 0.595 (0.361) 0.945 [2.157, 0.181]

PAT →technique →slope 0.149 (0.031) 4.184 (1.554) 0.623 [1.207, 0.153] 0.072 (0.035) 4.184 (1.554) 0.301 [0.740, 0.004]

IIP as outcome

PAT →dose →slope 1.923 (0.188) 0.016 (0.019) 0.031 [0.104, 0.041] 1.576 (0.217) 0.016 (0.019) 0.025 [0.087, 0.034]

PAT →technique →slope 0.148 (0.031) 0.043 (0.132) 0.006 [0.033, 0.047] 0.072 (0.034) 0.043 (0.132) 0.003 [0.018, 0.027]

INTREX as outcome

PAT →dose →slope 1.933 (0.186) 3.371 (2.868) 6.516 [17.622, 4.406] 1.596 (0.210) 3.371 (2.868) 5.380 [14.895, 3.513]

PAT →technique →slope 0.147 (0.031) 38.895 (17.049) 5.718 [0.737, 11.865] 0.070 (0.037) 38.895 (17.049) 2.723 [0.231, 7.294]

Coefﬁcients of Path aequal Band SE of the β

,β

and β

parameters in Table 3. We reversed the sign of these coefﬁcients to clarify that they represent increases in mediating var-

iables due to psychoanalytic therapy (as compared with control conditions). Coefﬁcients of Path bequal Band SE of the γ

and γ

parameters in Table 4. Ninety-ﬁve per cent conﬁdence

intervals for the product of aand bwere computed with the Monte Carlo method using 100 000 resamples. Signiﬁcant products of aand bare printed in bold. PAT = psychoanalytic ther-

apy. Dose = number of sessions/100. Technique= average Fisher-ztransformed proﬁle correlations with the psychoanalytic Psychotherapy Process Q-Set prototype. Slope =rate of change

per year during follow-up. BDI = Beck Depression Inventory. IIP-C = Inventory of Interpersonal Problems—Circumplex. INTREX = introject afﬁliation at worst.

481Is It All about the Higher Dose?

as compared with CBT in terms of post-treatment inter-

personal problems and introject may be primarily due

to the higher dose.

Explaining Outcome Differences between

Psychoanalytic and Psychodynamic Therapies

An important limitation of the mediation analyses pre-

sented above is that participants were not randomly

assigned to psychoanalytic therapy versus CBT. Thus, we

repeated the mediation analyses using psychodynamic

therapy as the control condition to ensure full randomiza-

tion of the independent variable (Huber et al., 2013).

Regarding Path a, Table 3 shows that psychoanalytic

therapy comprised a considerably higher number of

sessions than psychodynamic therapy (β

) and psychoana-

lytic interventions were more typical for psychoanalytic

therapy than for psychodynamic therapy (β

; with the lim-

itation that the latter difference was only marginally signiﬁ-

cant when pre-treatment introject afﬁliation was controlled

for). Results for Path bin Table 4 were identical to the results

presented above, as these paths are the same irrespective of

the contrast used in the multicategorical independent vari-

able (Figure 1). Moreover, the direct effects of psychoana-

lytic therapy again were not signiﬁcant (γ

and γ

suggesting that the mediating variables were successful at

explaining the total effects. Finally, the 95% Monte Carlo

CIs for the indirect effects shown in the last column of Ta-

ble 5 suggest that psychoanalytic technique mediated the

differential long-term effect of psychoanalytic therapy on

depressive symptoms. The remaining indirect effects were

not signiﬁcant.

Exploratory Item-Level Analyses of the Indirect Effect

on Depressive Symptoms

After determining that the psychoanalytic techniques

mediated the differential long-term effect of psychoana-

lytic therapy on depressive symptoms, we needed to

determine which of the prototypical psychoanalytic tech-

niques were responsible for this ﬁnding. Thus, we con-

ducted exploratory mediation analyses at the PQS item

level to ‘unpack’the indirect effect of the PA prototype.

Speciﬁcally, we reran the mediation analyses 40 times,

each time replacing the PA prototype score with one of

the 20 most characteristic and 20 most uncharacteristic

PQS items (Ablon & Jones, 1998). This is an appropriate

exploratory strategy as the PA prototype score is

essentially a weighted average of all 100 PQS items with

higher (absolute) weights for items that are more salient

(i.e., characteristic or uncharacteristic). Three PQS items

describing typical features of psychoanalytic therapies

predicted symptom reduction after therapy (Path b).

These items referred to discussing sexual feelings and

experiences (Item 11), B=0.54, SE = 0.24, patients’

dreams or fantasies (Item 90), B=0.34, SE = 0.17, and

memories or reconstructions from infancy and childhood

(Item 91), B=0.36, SE = 0.16, all p’s<0.05. In addition,

the last two signiﬁcantly mediated the relative effect of

psychoanalytic therapy (as compared with CBT) on long-

term outcome, with ab =0.45, 95% CI [1.09, 0.01],

for dreams or fantasies, and ab =0.48, 95% CI [1.08,

0.05], for memories from infancy and childhood. PQS

items describing uncharacteristic features of psychoana-

lytic therapies did not predict change after therapy, nor

did they mediate the effect of psychoanalytic therapy.

DISCUSSION

In this study, we tested whether the differential long-term

effect of psychoanalytic therapy (i.e., high-dose LTPP)

reported by Huber, Zimmermann, et al. (2012) and Huber

et al. (2013) would be mediated by the number of sessions

and/or by the features of the respective psychoanalytic

technique. We found partial support for the mediating

effect of psychoanalytic technique: psychoanalytic ther-

apy predicted a more pronounced application of psy-

choanalytic techniques, which in turn predicted more

favourable trajectories of depressive symptoms during

the 3-year follow-up period. Exploratory item analyses

suggested that the effect on depressive symptoms was

primarily driven by the therapist’s exploration of the

patient’s fantasy life and discussions of the patient’s early

memories. In addition, there was some indication for a

similar indirect effect on the improvement of introject

afﬁliation during follow-up, albeit this effect was signiﬁ-

cant only when using CBT as the comparison group.

Moreover, we also found support for a mediating effect

of dose on outcome: psychoanalytic therapy was associ-

ated with a strikingly higher number of sessions, which

in turn predicted fewer interpersonal problems and more

afﬁliative introjects directly after treatment. Thus, our

ﬁndings were in line with the prediction that both psycho-

analytic techniques and a higher dose would be involved

in establishing the differential effectiveness of psychoana-

lytic therapy: whereas the higher dose seems to be

For the sake of completeness, we also analysed differences between

psychoanalytic therapy and CBT in cognitive–behavioural technique

(Path a) and included the CBT prototype as a Level-2 predictor in

Equation 2 (Path b). Results for Path asuggested that cognitive–

behavioural technique was more salient in CBT sessions than in psy-

choanalytic therapy sessions, B= 0.23, SE = 0.03, p<0.001. Results for

Path bsuggested that the application of cognitive–behavioural tech-

nique was positively associated with post-treatment depressive symp-

toms, B= 13.38, SE = 6.40, p<0.05, but unrelated to long-term

outcome and other outcome measures. Note that controlling for

cognitive–behavioural technique in Equation 2 did not substantially

change the results of the effects of the other mediating variables.

482 J. Zimmermann et al.

responsible for the fact that psychoanalytic therapy is

effective in helping patients to reduce their interpersonal

problems and improve their self-love during treatment,

psychoanalytic techniques such as explorations of pa-

tients’fantasy lives and discussions of patients’early

memories seem to facilitate sustained change after

therapy.

Our ﬁndings regarding the positive effect of psychoana-

lytic technique on psychotherapeutic outcome are in line

with a number of previous process–outcome studies that

focused on short-term treatments for depression (Ablon

& Jones, 1998; Gaston et al., 1998; Hilsenroth et al., 2003).

However, our ﬁndings extend such previous studies in

several ways: ﬁrst, these previous studies assessed the

association between psychodynamic–interpersonal or

psychoanalytic technique and outcome within speciﬁc

treatments and did not test for the mediating role of these

techniques in facilitating the differential effectiveness of

treatments as we did. Thus, our study represents a major

step forward according to current recommendations for

research on psychotherapeutic processes (Kazdin, 2007).

Second, in the current study, psychoanalytic techniques

were assessed using three separate sessions per treatment,

which clearly improved the reliability and representative-

ness of process measurement as compared with assess-

ments using only a single session per treatment. Third,

previous studies focused on the immediate effects on

psychotherapeutic outcome, whereas our study is the ﬁrst

to include assessments during a 3-year follow-up period.

Finally, our ﬁndings show that psychoanalytic techniques

are also relevant in long-term treatments, even when

controlling for the higher number of sessions. This has

important implications for interpreting the growing evi-

dence on the effectiveness of LTPP (de Maat et al., 2013;

Leichsenring & Rabung, 2008, 2011; Leichsenring, Abbass,

et al. 2013; Smit et al., 2012). Our study provides the ﬁrst

evidence that intensive psychodynamic treatments such

as psychoanalytic therapy are effective not only because

they are intensive but also because they are psychoanalytic.

The exploratory ﬁndings on the PQS item level should

be interpreted with caution as the speciﬁc pattern of

results may have been due to chance. Nevertheless, it

seems striking that Slavin-Mulford et al. (2011) identiﬁed

very similar items as relevant for signiﬁcant clinical

change, namely focusing on wishes, fantasies, dreams

and early memories. The common denominator of these

items seems to be that they do not represent therapist

interventions in a strict sense but rather the contents of

the sessions. From a methodological point of view, this

could be due to the fact that ratings of content are more

reliable than ratings of procedural aspects. A more clinical

explanation would be that psychoanalytic therapies

provide time, space and an ‘interpsychic’dimension

(Bolognini, 2004) for reverie and containment so that

patients are more likely to say things that are hard to say

and that are almost never said to anybody else. In turn,

verbalizing one’s own dreams, fantasies and early

memories might help a person to reorganize maladaptive

representations of oneself and others and lead to sus-

tained change even after treatment termination. Of course,

this interpretation remains speculative. Further studies are

needed to replicate the relevance of these speciﬁc tech-

niques, and, even more important, to explore how they

produce sustained change in patients.

Another implication of our ﬁndings is that dose and

technique have different effects on different outcomes:

whereas improvements in depression and hostility to-

wards the self after treatment were related to psycho-

analytic technique, improvements in interpersonal

problems and hostility towards the self during treatment

were linked to the number of sessions. The former ﬁnding

suggests that sustained therapeutic change may depend,

at least to some extent, on the application of speciﬁc

strategies that will foster change in the underpinning

intrapsychic structures (e.g., the development of a ‘self-

analytic function’; Falkenström, Grant, Broberg, &

Sandell, 2007) that will remain stable even when the

therapist is not actually present. The latter ﬁnding

suggests that changes in interpersonal problems and self-

acceptance may simply require that enough time be spent

in a positive and caring therapeutic relationship. This

latter ﬁnding is also in line with the meta-analysis by

Leichsenring and Rabung (2011) who reported that treat-

ment duration was more highly correlated with social

functioning than with psychiatric symptoms, and with

dose–response studies reporting that recovery from acute

distress typically evolves more quickly than recovery from

impairments in social or personality functioning (e.g.,

Kopta et al., 1994).

The present study has several limitations. First, neither

the treatment nor the mediator variables were fully ran-

domized, and thus, causal inferences are questionable

(cf. Bullock, Green, & Ha, 2010; Imai, Keele, & Tingley,

2010). That is, we cannot rule out the possibility that our

ﬁndings were inﬂuenced or biased by other variables that

were not included in our statistical models (e.g., thera-

peutic alliance, therapist competence or early response).

However, patients were at least fully randomized to

psychoanalytic versus psychodynamic therapy, sug-

gesting that the internal validity of the effects pertaining

to these analyses is somewhat higher. Moreover, pre-

dicting changes after treatment clearly ensured that

treatment allocation, mediators and outcome assessment

followed each other in time (i.e., effects of changes after

treatment on mediators were causally impossible). This

suggests that the reported effects of psychoanalytic tech-

nique on outcome during follow-up are methodologically

more convincing than the effects of dose on outcome at

treatment termination, as the latter effects may also work

the other way around (e.g., therapists might prefer to

483Is It All about the Higher Dose?

continue therapy with patients who have reduced inter-

personal problems because they regard these sessions as

more ‘comfortable’).

Second, the use of middle sessions for technique ratings

has drawbacks. One issue is that the application of tech-

niques in middle sessions might be inﬂuenced by prior

symptom change. For example, one might argue that in

therapies with patients who show an early response, it is

easier for therapists to conduct psychoanalytic interven-

tions, and thus, the association between psychoanalytic

technique and long-term outcome might be spurious due

to the inﬂuence of third variables that are responsible for

the early response. However, from a conceptual perspec-

tive, this argument would also apply if early sessions were

used for the technique ratings as only the randomization of

technique warrants causal inferences (e.g., Høglend et al.,

2006). From an empirical perspective, this interpretation

seems rather unlikely because if the application of psycho-

analytic technique were to be associated with changes in

outcome measures during the ﬁrst half of the treatment,

one would also expect that it would be associated with

changes during the full treatment, and this was apparently

not the case (i.e., the coefﬁcients for γ

were generally not

signiﬁcant). A further issue regarding the use of middle

sessions is that the (arithmetic) middle varies across

treatments and is completely confounded with dose. For

example, in CBT, technique was usually assessed in the

25th session, whereas in psychoanalytic therapy, it was,

on average, assessed in the 120th session. Given that

patients are more likely to bring intimate content to

session 120 than to session 25, our measure of psychoana-

lytic technique might simply be a measure of the ‘depth’

or ‘progress’of treatment and not of the psychoanalytic

technique per se. However, if this alternative interpretation

were valid, one would expect higher PA prototype scores

in middle sessions that occur later (e.g., in the 120th as

compared with the 25th session) irrespective of the

speciﬁc treatment modality. In other words, when con-

sidering sessions from the three treatment conditions

separately, the session numbers of the middle sessions in

which psychoanalytic technique was assessed (i.e., half

of the dose) should be positively correlated with the actual

amount of psychoanalytic technique administered. Obvi-

ously, this was not the case, as dose and technique were

uncorrelated when controlling for treatment condition.

In any case, further investigations of a broader and more

balanced sample of sessions from the MPS is clearly

needed and currently underway (e.g., early sessions after

aﬁxed number of sessions or sessions with ‘signiﬁcant

events’based on therapists’process ratings).

Third, the naturalistic setting of this study poses further

threats to the validity of our conclusions. For example,

due to the lack of manuals and quality checks, we cannot

be sure that therapists applied the techniques equally com-

petently in the different treatment conditions. Moreover,

although our ﬁndings suggest that treatment adherence

was adequate (e.g., the application of psychoanalytic

technique was most salient in the psychoanalytic therapy

sessions), the PA prototype scores were generally lower

than in former studies (Ablon & Jones, 2005). One

explanation for this could be that the therapists in the

MPS who delivered the psychoanalytic and psychody-

namic treatments adhered less stringently to the principles

of psychoanalytic therapy than their American colleagues

in the studies analysed by Ablon and Jones (1998, 2005).

Another explanation could be that raters, as they had no

psychotherapeutic expertise and showed rather modest

inter-rater reliability, might have systematically over-

looked the application of some psychoanalytic techniques.

Obviously, further research is needed to investigate the

effects of psychoanalytic techniques on outcome using

highly controlled efﬁcacy designs.

Fourth, the majority of authors involved in conducting

this study had an allegiance to psychoanalytic therapy

(e.g., according to prior publications, clinical training and

professional afﬁliations). As researcher allegiance carries

a risk of bias in psychotherapy research (Munder et al.,

2013), we cannot rule out that this kind of bias inﬂuenced

our results. Fifth, sample size was probably sufﬁcient only

for detecting indirect effects of a medium effect size (Fritz

& MacKinnon, 2007). For example, dose might indeed

have a small effect on long-term change in depressive

symptoms, as the respective effect γ

just missed statistical

signiﬁcance, with p= 0.10. Finally, outcome was assessed

using self-report questionnaires, and technique was

assessed using observer reports. Additional observer-rated

outcome and patient-rated process variables would have

provided a more holistic and in-depth view into ongoing

changes (Löfﬂer-Stastka, 2011).

Despite these limitations, the present study provides an

important step towards empirically unravelling the

effectiveness of LTPP. Our results tentatively suggest that

It should be noted that a causal interpretation of the mediating ef-

fect of dose could also be questioned on conceptual grounds. One as-

sumption of causal mediation analysis is that the treatment must be

independent of all potential values of the mediating variable (i.e., that

each treatment in theory can receive any level of the mediator; Imai

et al., 2010). This assumption might not hold for number of sessions

as treatments such as ‘low-dose psychoanalytic therapy’or ‘high-

dose CBT’might seem conceptually unreasonable. In other words,

number of sessions might be conceived of as a deﬁning feature of

the treatments as opposed to an intermediate variable that is manip-

ulated by the treatments. However, although patients received, on av-

erage, strikingly higher numbers of sessions in psychoanalytic

therapy than in psychodynamic therapy and CBT, the distributions

of the three groups clearly overlapped. That is, from an empirical per-

spective, number of sessions cannot be simply regarded as a proxy for

treatment. Moreover, even when one disregards dose as a potential

causal mediator, our main ﬁnding that psychoanalytic techniques

predict long-term outcome above and beyond dose would still be

valid.

484 J. Zimmermann et al.

psychoanalytic techniques play a more prominent role in

establishing sustained changes in patients with major

depression than has been previously thought. That is,

intensive psychodynamic treatments such as psychoana-

lytic therapy seem to be effective not only because they

are intensive but also because they are psychoanalytic.We

are aware of the fact that this conclusion is highly prelim-

inary and that further studies replicating our ﬁndings are

urgently needed. In this regard, we hope that our study

will encourage researchers to investigate psychotherapeu-

tic processes in long-term treatments and to gauge

whether their effects outweigh their higher costs across

long-term follow-up periods (e.g., Benecke et al., 2014;

Beutel et al., 2012). Although such studies are expensive,

time-consuming and at odds with the current publish-or-

perish practice, they may be the only way to empirically

address some of the most troubling questions of our ﬁeld:

the how and when of sustained therapeutic change.

AKNOWLEDGEMENTS

This study was supported by grants from the Research

Advisory Board of the International Psychoanalytic

Association (IPA) and from the Dr. Zita und T. V. Steger

Foundation. The authors thank Simone Wieser for her

help with digitizing the audio tapes, Christine Rohm

and Christian Hau for conducting the PQS ratings, and

Jane Zagorski for proofreading the article.

REFERENCES

Ablon, J. S., & Jones, E. E. (1998). How expert clinicians’proto-

types of an ideal treatment correlate with outcome in psycho-

dynamic and cognitive–behavior therapy. Psychotherapy

Research,8,71–83.

Ablon, J. S., & Jones, E. E. (2002). Validity of controlled clinical tri-

als of psychotherapy: Findings from the NIMH Treatment of

Depression Collaborative Research Program. American Journal

of Psychiatry,159, 775–783.

Ablon, J. S., & Jones, E. E. (2005). On analytic process. Journal of

the American Psychoanalytic Association,53, 541–568.

Ablon, J. S., Levy, R. A., & Katzenstein, T. (2006). Beyond brand

names of psychotherapy: Identifying empirically supported

change processes. Psychotherapy: Theory, Research, Practice,

Training,43, 216–231.

Albani, C., Blaser, G., Jacobs, U., Jones, E., Geyer, M., & Kächele, H.

(2000). Die Methode des ‘Psychotherapie-ProzeβQ-Sort’[The

Psychotherapy Process Q-Sort method]. Zeitschrift für Klinische

Psychologie, Psychiatrie und Psychotherapie,48,151–171.

Alden, L. E., Wiggins, J. S., & Pincus, A. L. (1990). Construction of

circumplex scales for the Inventory of Interpersonal Problems.

Journal of Personality Assessment,55, 521–536.

Anderson, E. M., & Lambert, M. J. (2001). A survival analysis of

clinically signiﬁcant change in outpatient psychotherapy. Jour-

nal of Clinical Psychology,57, 875–888.

Barkham, M., Rees, A., Stiles, W. B., Shapiro, D. A., Hardy, G. E.,

& Reynolds, S. (1996). Dose–effect relations in time-limited

psychotherapy for depression. Journal of Consulting and Clinical

Psychology,64, 927–935.

Beck, A. T., Steer, R. A. & Garbin, M. G. (1988). Psychometric

properties of the Beck Depression Inventory: Twenty-ﬁve years

of evaluation. Clinical Psychology Review,8,77–100.

Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J.

(1961). An inventory for measuring depression. Archives of

General Psychiatry,4, 561–571.

Bedics, J. D., Atkins, D. C., Comtois, K. A., & Linehan, M. M.

(2012). Treatment differences in the therapeutic relationship

andintrojectduringa2-yearrandomizedcontrolledtrial

of dialectical behavior therapy versus nonbehavioral

psychotherapy experts for borderline personality disorder.

Journal of Consulting and Clinical Psychology,80,66–77.

Beenen, F., & Stoker, J. (1996). Periodical rating scale for psychoan-

alytic treatment. Amsterdam: Netherlands Psychoanalytic

Institute.

Benecke, C., Huber, D., Staats, H., Zimmermann, J., Deserno, H.,

Wiegand-Grefe, S., & Schauenburg, H. (2014). The APS study:

A comparison of psychoanalytic therapy and cognitive-behavior

therapy for anxiety and personality disorders. Presentation of

the RCT study design. Manuscript in preparation.

Benjamin, L. S. (1974). Structural analysis of social behavior.

Psychological Review,81, 392–425.

Benjamin, L. S. (1983). The INTREX user’s manual. Part I and II.

Intrex Interpersonal Institute, Inc., P.O. Box 55218, Madison,

WI 53705, Madison.

Beutel, M. E., Leuzinger-Bohleber, M., Rüger, B., Bahrke, U.,

Negele, A., Haselbacher, A., …Hautzinger, M. (2012).

Psychoanalytic and cognitive–behavior therapy of chronic

depression: study protocol for a randomized controlled

trial. Trials,13,117.

Blagys, M. D., & Hilsenroth, M. J. (2000). Distinctive feature of

short-term psychodynamic–interpersonal psychotherapy: A

review of the comparative psychotherapy process literature.

Clinical Psychology: Science and Practice,7, 167–188.

Bolognini, S. (2004). Intrapsychic–interpsychic. International Jour-

nal of Psychoanalysis,85, 337–357.

Bullock, J. G., Green, D. P., & Ha, S. E. (2010). Yes, but what’s the

mechanism? (Don’t expect an easy answer). Journal of Perso-

nality and Social Psychology,98, 550–558.

Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for

evaluating normed and standardized assessment instruments

in psychology. Psychological Assessment,6, 284–290.

Cuijpers, P., van Straten, A., Andersson, G., & van Oppen, P.

(2008). Psychotherapy for depression in adults: A meta-

analysis of comparative outcome studies. Journal of Consulting

and Clinical Psychology,76, 909–922.

de Maat, S., de Jonghe, F., de Kraker, R., Leichsenring, F., Abbass, A.,

Luyten, P., …Dekker, J. (2013). The current state of the empirical

evidence for psychoanalysis: A meta-analytic approach. Harvard

Review of Psychiatry,21, 107–137.

Dekker, J., Molenaar, P. J., Kool, S., van Aalst, G., Peen, J., &

Jonghe, F. de. (2005). Dose–effect relations in time-limited

combined psycho-pharmacological treatment for depression.

Psychological Medicine: A Journal of Research in Psychiatry and

the Allied Sciences,35,47–58.

Driessen, E., Cuijpers, P., Maat, S. C. de, Abbass, A. A., Jonghe, F.

de, & Dekker, J. J. (2010). The efﬁcacy of short-term psychody-

namic psychotherapy for depression: A meta-analysis. Clinical

Psychology Review,30,25–36.

485Is It All about the Higher Dose?

Falkenström, F., Grant, J., Broberg, J., & Sandell, R. (2007). Self-

analysis and post-termination improvement after psychoanal-

ysis and long-term psychotherapy. Journal of the American

Psychoanalytic Association,55, 629–674.

Fonagy, P., & Kächele, H. (2009). Psychoanalysis and other

long-term dynamic psychotherapies. In M.-G. Gelder, J. J.

Lopez-Ibor, & N. Andreasen (Eds.), New Oxford textbook of

psychiatry (vol. 2, pp. 1337–1349). Oxford: Oxford Uni-

versity Press.

Fritz, M. S., & MacKinnon, D. P. (2007). Required sample size to

detect the mediated effect. Psychological Science,18, 233–239.

Gabbard, G. O. (2004). Long-term psychodynamic psychotherapy.

Washington, DC: American Psychiatric Publishing.

Gabbard, G. O., & Horowitz, M. J. (2009). Insight, transference

interpretation, and therapeutic change in the dynamic psycho-

therapy of borderline personality disorder. American Journal of

Psychiatry,166, 517–521.

Gaston, L., Thompson, L., Gallagher, D., Cournoyer, L.-G., &

Gagnon, R. (1998). Alliance, technique, and their interactions

in predicting outcome of behavioral, cognitive, and brief

dynamic therapy. Psychotherapy Research,8, 190–209.

Gerber, A. J., Kocsis, J. H., Milrod, B. L., Roose, S. P., Barber, J. P.,

Thase, M. E., …Leon, A. C. (2011). A quality-based review of

randomized controlled trials of psychodynamic psychother-

apy. American Journal of Psychiatry,168,19–28.

Gibbons, M. B. C., Crits-Christoph, P., & Hearon, B. (2008). The

empirical status of psychodynamic therapies. Annual Review

of Clinical Psychology,4,93–108.

Hautzinger, M., Bailer, M., Worall, H., & Keller, F. (1995). Beck-De-

pressions-Inventar (BDI): Testhandbuch (2. Auﬂ.) [Beck

Depression Inventory]. Bern: Hans Huber.

Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical

mediation analysis in the new millennium. Communication

Monographs,76, 408–420.

Hayes, A. F., & Preacher, K. J. (2011). Indirect and direct effects of

a multicategorical causal agent in statistical mediation analy-

sis. Unpublished manuscript.

Henry, W. P., Strupp, H. H., Schacht, T. E., & Gaston, L. (1994).

Psychodynamic approaches. In A. E. Bergin & S. L. Garﬁeld

(Eds.), Handbook of psychotherapy and behavior change

(pp. 467–508). New York: Wiley.

Hilsenroth, M. J., Ackerman, S. J., Blagys, M. D., Baity, M. R., &

Mooney, M. A. (2003). Short-term psychodynamic psycho-

therapy for depression: An examination of statistical, clinically

signiﬁcant, and technique-speciﬁc change. Journal of Nervous

and Mental Disease,191, 349–357.

Hilsenroth, M., Blagys, M., Ackerman, S., Bonge, D., & Blais,

M. (2005). Measuring psychodynamic–interpersonal and

cognitive–behavioral techniques: Development of the Com-

parative Psychotherapy Process Scale. Psychotherapy,42,

340–356.

Høglend, P., Amlo, S., Marble, A., Bøgwald, K.-P., Sørbye, O.,

Sjaastad, M. C., & Heyerdahl, O. (2006). Analysis of the

patient–therapist relationship in dynamic psychotherapy: An

experimental study of transference interpretations. American

Journal of Psychiatry,163, 1739–1746.

Hollon, S. D., & Beck, A. T. (2004). Cognitive and cognitive

behavioral therapies. In M. J. Lambert (Ed.), Bergin and

Garﬁeld’s handbook of psychotherapy research und behavior change

(5th ed., pp. 447–492). New York: Wiley.

Horowitz, L. M., Strauß, B., Kordy, H. (2000): Inventar zur

Erfassung Interpersoneller Probleme [Inventory of Interpersonal

Problems]. Weinheim: Beltz.

Howard, K. I., Kopta, S. M., Krause, M. S., & Orlinsky, D. E.

(1986). The dose–effect relationship in psychotherapy.

American Psychologist,41, 159–164.

Huber, D., Henrich, G., Clarkin, J., & Klug, G. (2013). Psychoana-

lytic versus psychodynamic therapy for depression: A three-

year follow-up study. Psychiatry,76, 132–149.

Huber, D., Henrich, G., Gastner, J., & Klug, G. (2012). Must all

have prices? The Munich psychotherapy study. In: R. A. Levy,

J. S. Ablon, & H. Kächele (Eds.), Psychodynamic psychotherapy

research. Evidence-based practice and practice-based evidence

(pp. 51–69). New York: Humana Press.

Huber, D., Zimmermann, J., Henrich, G., & Klug, G. (2012). Com-

parison of cognitive–behaviour therapy with psychoanalytic

and psychodynamic therapy for depressed patients: A three-

year follow-up study. Zeitschrift für Psychosomatische Medizin

und Psychotherapie,58, 299–316.

Hughes, J., & Barkham, M. (2005). Scoping the Inventory of Inter-

personal Problems, its derivations and short forms: 1988–2004.

Clinical Psychology and Psychotherapy,12, 475–496.

Imai, K., Keele, L., & Tingley, D. (2010). A general approach to

causal mediation analysis. Psychological Methods,15, 309–334.

Jones, E. E. (1985). Manual for the psychotherapy Q sort. Unpub-

lished manuscript, University of California, Berkeley.

Jones, E. E., & Pulos, S. M. (1993). Comparing the process in

psychodynamic and cognitive–behavioral therapies. Journal of

Consulting and Clinical Psychology,61, 306–316.

Jones, E. E., Cumming, J. D., & Horowitz, M. J. (1988). Another

look at the nonspeciﬁc hypothesis of therapeutic effectiveness.

Journal of Consulting and Clinical Psychology,56,48–55.

Kazdin, A. E. (2007). Mediators and mechanisms of change in

psychotherapy research. Annual Review of Clinical Psychology,

3,1–17.

Kopta, S. M., Howard, K. I., Lowry, J. L., & Beutler, L. E. (1994).

Patterns of symptomatic recovery in psychotherapy. Journal of

Consulting and Clinical Psychology,62, 1009–1016.

Krause, R. (2009). Psychodynamische Intervention [Psychody-

namic intervention]. In: M. Hautzinger, & P. Pauli (Eds.),

Psychotherapeutische Methoden (pp. 161–221). Göttingen:

Hogrefe.

Krull, J. L., & MacKinnon, D. P. (2001). Multilevel modeling of

individual and group level mediated effects. Multivariate

Behavioral Research,36, 249–277.

Lambert, M. J., & Ogles, B. M. (2004). The efﬁcacy and effective-

ness of psychotherapy. In M. J. Lambert (Ed.), Bergin and

Garﬁeld’s handbook of psychotherapy and behavior change

(5th ed., pp. 139–193). New York: Wiley.

Lambert, M. J., Hansen, N. B., & Finch, A. E. (2001). Patient-

focused research: Using patient outcome data to enhance treat-

ment effects. Journal of Consulting and Clinical Psychology,69,

159–172.

Leichsenring, F. (2001). Comparative effects of short-term psy-

chodynamic psychotherapy and cognitive–behavioral therapy

in depression. Clinical Psychology Review,21, 401–419.

Leichsenring, F., Abbass, A., Luyten, P., Hilsenroth, M., &

Rabung, S. (2013). The emerging evidence for long-term psy-

chodynamic therapy. Psychodynamic Psychiatry,41, 361–384.

Leichsenring, F., & Rabung, S. (2008). Effectiveness of long-

term psychodynamic psychotherapy: A meta-analysis.

JAMA,theJournaloftheAmericanMedicalAssociation,300,

1551–1565.

Leichsenring, F., & Rabung, S. (2011). Long-term psychodynamic

psychotherapy in complex mental disorders: Update of a meta-

analysis. British Journal of Psychiatry,199,15–22.

486 J. Zimmermann et al.

Leichsenring, F., Salzer, S., Beutel, M. E., Herpertz, S., Hiller, W.,

Hoyer, J., …Leibing, E. (2013). Psychodynamic therapy and

cognitive–behavioral therapy in social anxiety disorder: A

multicenter randomized controlled trial. American Journal of

Psychiatry,170, 759–767.

Löfﬂer-Stastka, H. (2011). Q-sort-Methoden in Diagnostik und

Therapieverlaufsbeobachtung [Use of Q-sort assessment

methods for diagnostic purposes and in treatment evaluation].

Zeitschrift für Psychosomatische Medizin und Psychotherapie,57,

319–324.

Long, J. S., & Ervin, L. H. (2000). Using heteroscedasticity consistent

standard errors in the linear regression model. The American

Statistician,54,217–224.

MacKinnon, D. P. (2008). Introduction to statistical mediation analy-

sis. New York, NY: Erlbaum.

MacKinnon,D.P.,Lockwood,C.M.,&Williams,J.(2004).Conﬁdence

limits for the indirect effect: Distribution of the product and

resampling methods. Multivariate Behavioral Research,39,41–62.

Monsen, J. T., von der Lippe, A. L., Havik, O. E., Halvorsen, M.

S., & Eilertsen, D. E. (2007). Validation of the SASB Introject

Surface in a Norwegian clinical and nonclinical sample. Journal

of Personality Assessment,88, 235–245.

Munder, T., Brütsch, O., Leonhart, R., Gerger, H., & Barth, J. (2013).

Researcher allegiance in psychotherapy outcome research: An

overview of reviews. Clinical Psychology Review,33,501–511.

Owen,J.,&Hilsenroth,M.J.(2011).Interaction between alliance and

technique in predicting patient outcome during psychodynamic

psychotherapy. Journal of Nervous and Mental Disease,199,384–389.

Pincus, A. L., Gurtman, M. B., & Ruiz, M. A. (1998). Structural

Analysis of Social Behavior (SASB): Circumplex analyses and

structural relations with the interpersonal circle and ﬁve-

factor model of personality. Journal of Personality and Social

Psychology,74, 1629–1645.

Pincus, A. L., Newes, S. L., Dickinson, K. A., & Ruiz, M. A. (1998).

A comparison of three indexes to assess the dimensions of

Structural Analysis of Social Behavior. Journal of Personality

Assessment,70, 145–170.

Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for

estimating indirect effects in simple mediation models. Behavior

Research Methods, Instruments, & Computers,36,717–731.

Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling

strategies for assessing and comparing indirect effects in multi-

ple mediator models. Behavior Research Methods,40, 879–891.

Preacher, K. J., & Selig, J. P. (2012). Advantages of Monte Carlo

conﬁdence intervals for indirect effects. Communication Methods

and Measures,6,77–98.

R. Core Team. (2013). R: A language and environment for

statistical computing. Vienna: R Foundation for Statistical

Computing. Retrieved from http://www.R-project.org.

Rüger, U., Dahm, A., & Kallinke, D. (2003). Faber Haarstrick

Kommentar Psychotherapie-Richtlinien, 6. Auﬂ.[TheFaber–

Haarstrick commentary on psychotherapy guidelines].

München: Urban & Fischer.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of

the state of the art. Psychological Methods,7, 147–177.

Selig, J. P., & Preacher, K. J. (2008). Monte Carlo method for

assessing mediation: An interactive tool for creating conﬁ-

dence intervals for indirect effects [Computer software].

Available from http://quantpsy.org/.

Shedler, J. (2010). The efﬁcacy of psychodynamic psychotherapy.

American Psychologist,65,98–109.

Slavin-Mulford, J., Hilsenroth, M., Weinberger, J., & Gold, J.

(2011). Therapeutic interventions related to outcome in psycho-

dynamic psychotherapy for anxiety disorder patients. Journal

of Nervous and Mental Disease,199, 214–221.

Smit, Y., Huibers, M. J., Ioannidis, J. P., van Dyck, R., van Tilburg,

W., & Arntz, A. (2012). The effectiveness of long-term psycho-

analytic psychotherapy—A meta-analysis of randomized con-

trolled trials. Clinical Psychology Review,32,81–92.

Thorpe, K. E., Zwarenstein, M., Oxman, A. D., Treweek, S., Furberg,

C. D., Altman, D. G., …Chalkidou, K. (2009). A pragmatic-

explanatory continuum indicator summary (PRECIS): A tool

to help trial designers. Journal of Clinical Epidemiology,62,

464–475.

Tracey, T. J. G., Rounds, J. B., & Gurtman, M. B. (1996). Examina-

tion of the general factor with the interpersonal circumplex

structure: Application to the Inventory of Interpersonal

Problems. Multivariate Behavior Research,31, 441–466.

Tress, W. (1993). INTREX-Kurzform-Fragebogen: Formulare,

Itemformulierungen und Fragestellungen [INTREX short form

questionnaire: Forms, item wordings, and questions]. In W. Tress

(Ed.), SASB. Die strukturale Analyse sozialen Verhaltens

(pp. 259–263). Heidelberg: Asange.

Wampold, B. E., & Serlin, R. C. (2000). The consequence of

ignoring a nested factor on measures of effect size in analysis

of variance. Psychological Methods,5, 425–433.

487Is It All about the Higher Dose?

ResearchGate has not been able to resolve any citations for this publication.

A Comparison of Psychoanalytic Therapy and Cognitive Behavioral Therapy for Anxiety (Panic/Agoraphobia) and Personality Disorders (APD Study): Presentation of the RCT Study Design

Article

Full-text available

Sep 2016
Z PSYCHOSOM MED PSYC

Objectives: Anxiety disorders, most notably panic disorders and agoraphobia, are common mental disorders, and there is a high comorbidity with personality disorders. Randomized controlled trails addressing this highly relevant group of patients are missing. Design: The multicenter Anxiety and Personality Disorders (APD) study investigates 200 patients with panic disorder and/or agoraphobia with comorbid personality disorder in a randomized control-group comparison of psychoanalytic therapy (PT) and cognitive behavioral therapy (CBT), including 100 patients in each group. Each patient will be examined over a period of six years, regardless of the duration of the individual treatment. The main issues that are addressed in this study are the comparison of the efficacy of PT and CBT in this special patient population, the comparison of the sustainability of the effects of PT and CBT, the comparison of the long-term cost-benefit-ratios of PT and CBT as well as the investigation of prescriptive patient characteristics for individualized treatment recommendations (differential indication). Discussion: The APD study compares efficacy, sustainability, and cost-benefit-ratios of CBT and PT for anxiety plus personality disorders in a randomized controlled trail. The study design meets the requirements for an efficacy study for PT, which were recently defined. Trial registration: Current Controlled Trials ISRCTN12449681.

Confidence Limits for the Indirect Effect: Distribution of the Product and Resampling Methods

Article

Full-text available

Jan 2004
MULTIVAR BEHAV RES

The most commonly used method to test an indirect effect is to divide the estimate of the indirect effect by its standard error and compare the resulting z statistic with a critical value from the standard normal distribution. Confidence limits for the indirect effect are also typically based on critical values from the standard normal distribution. This article uses a simulation study to demonstrate that confidence limits are imbalanced because the distribution of the indirect effect is normal only in special cases. Two alternatives for improving the performance of confidence limits for the indirect effect are evaluated: (a) a method based on the distribution of the product of two normal random variables, and (b) resampling methods. In Study 1, confidence limits based on the distribution of the product are more accurate than methods based on an assumed normal distribution but confidence limits are still imbalanced. Study 2 demonstrates that more accurate confidence limits are obtained using resampling methods, with the bias-corrected bootstrap the best method overall.

The International Journal of Psychoanalysis as the voice of psychoanalysis

Chapter

Feb 2021

Rachel B. Blass

Psychoanalysis and other long-term dynamic psychotherapies

Chapter

Feb 2012

Psychoanalysis is hardly a practical treatment alternative for the twenty-first century. The principles derived from this treatment, however, have powerfully influenced other psychotherapeutic approaches, whether long-term or short-term therapy or psychiatric care more generally, particularly in the United States. At the time of its invention, it was the unique effective psychosocial treatment method for psychiatric disorder which offered a genuine alternative to the sometimes barbaric and generally ineffective treatment methods available. Not surprisingly, its proponents adopted an almost religious zeal in defending its value against alternative approaches. While understandable, such an attitude has no place in the sophisticated evidence base underpinning multi-agency service planning. Psychoanalytic clinicians face a challenge in identifying their niche in the complex mental health care delivery systems of the twenty-first century.

Psychometric properties of the Beck depression inventory

Article

Jan 1988
CLIN PSYCHOL REV

The 'Psychotherapy Process Q-Sort' method

Article

Jan 2000

The article provides an overview of current research with the 'Psychotherapy Process Q-Sort' by Enrico Jones and presents the application of a German version developed by the authors. The method could be used reliably and proved a proper instrument for describing psychotherapeutic processes. Comparison of sessions from the beginning and termination phases of a psychoanalytic therapy demonstrated clinically relevant differences between the two phases of treatment.

More on evidence-based practice in psychotherapy research

Article

Nov 2004

Nada Dimcovic

I discuss research and clinical reality from the point of view of a psychologist with a special interest in eating disorders.

Psychodynamic psychotherapy

Article

Sep 2008

The focus of psychodynamic psychotherapy is understanding how past experiences influence present behaviours. Psychodynamic psychotherapy can be divided into two subtypes: brief or time-limited therapy and long-term or open-ended therapy. The core concepts of psychodynamic psychotherapy include the unconscious, transference, counter transference, resistance, multiple functions, authenticity and a developmental perspective. Long-term psychodynamic psychotherapy is derived from psychoanalysis and the work of Sigmund Freud. Freud's original focus was on the topographic model, which involved a three-level hierarchy of conscious, preconscious and unconscious domains. The technique of dynamic therapy has evolved from a largely silent and nondirective role for the therapist to one in which the therapist is lively and interactive. Dynamic therapy often requires a considerable length of time because many patients will tenaciously hold on to self-defeating patterns and problematic object relationships. Patients must repeat the old patterns again and again to understand them in different contexts.

Die Methode des "Psychotherapie-Prozeß Q-Sort"

Article

Feb 2000

Psychodynamic approaches

Article

All just a Question of the larger dose? Why analytic Psychotherapy in depressive Disorders is effective in the long term

Abstract and Figures

Recommended publications

Impasse with routine psychiatric treatment: progress with analytic psychotherapy in a case of neurot...

Psychoterapia oparta na analizie funkcjonalnej (Functional Analytic Psychotherapy, FAP): kliniczne z...

Causal beliefs amongst families in therapy: Attributions at the group level

Éléments cliniques sur l'indication des psychothérapies brèves focales, ou focalisées, de l'enfant C...