Background
According to estimates of the World Health Organization, 4.4% of the global population suffered from depression in 2015 [
1]. Given the high prevalence of major depressive disorder and its severe impact on the functioning of patients who suffer from it, it is not surprising that depression is one of the three leading causes of the global disease burden [
2]. Fortunately, several evidence-based pharmaco- and psychotherapeutic treatment options are available. Research suggests that most patients suffering from a psychiatric disorder prefer psychotherapy over medication [
3]. Numerous well-conducted studies have demonstrated the effectiveness of several kinds of psychotherapy for MDD [
4]. Although cognitive behavioural therapy (CBT) has been studied most [
5,
6], there is no robust evidence indicating that efficacy varies between CBT and other evidence-based types of psychotherapeutic treatment for MDD such as interpersonal psychotherapy, behavioural activation, problem-solving therapy and psychodynamic psychotherapy [
7‐
9]. The fact that the outcome estimates of different kinds of psychotherapy for MDD seem to be quite similar has led some people to argue that the effects of the different available psychotherapeutic treatment methods can mainly be attributed to common factors [
10‐
12]. Common or non-specific factors are thought to be universal in all psychotherapy rather than specific to the particular method used. On the other hand, there are strong arguments against this common-factor hypothesis. See for a comprehensive overview Cuijpers, Reijnders and Huibers [
13]. The presumed equivalence of different forms of psychotherapy therefore gives rise to an important debate about the possible working mechanisms underlying psychotherapy for depression.
The availability of so many different evidence-based psychotherapeutic treatments for depression that at least appear to yield comparable treatment results can nevertheless be viewed as a good thing. It means that there are several viable options available to which the therapist can resort if response is insufficient, particularly if one considers that both patients and therapists may have different preferences for particular types of treatment.
Worldwide, a substantial number of psychotherapists practise psychotherapy from a psychoanalytic perspective [
14,
15]. The first major objective of the present study is to add to the evidence base for Short-term Psychodynamic Supportive Psychotherapy (SPSP) [
16,
17]. SPSP is a specific form of psychodynamic psychotherapy that has been tested in several RCTs in the Netherlands [
18] and was found to be non-inferior to Cognitive Behavioural Therapy (CBT) for patients with major depressive disorder [
19]. The first part of the study consists of an RCT in which patients are randomly assigned to Short-term Psychodynamic Supportive Psychotherapy (SPSP) or Cognitive Behavioural Therapy (CBT).
The empirically established overall effectiveness of psychotherapy does not imply that all patients will benefit from treatment. A meta-analysis of the treatment outcomes of evidence-based forms of psychotherapy estimates that the number of patients that does not respond (as defined by an improvement of 50% or more on any given self-report depression-severity scale) is substantial (52%), while 57 to 59% do not achieve full remission after one treatment option [
4] (as defined by a cut-off score on a depression-severity scale [
20]). This implies an improvement in the treatment results in psychotherapy for depression is urgently needed. But where to begin? The development of new forms of psychotherapy over the years does not seem to have made treatment more effective [
21]. One cannot therefore expect this to be automatically changed by the introduction of SPSP as a new form of evidence-based treatment for depression. The proponents of the ‘common factor’ hypothesis would also recommend focusing more on researching and working on the optimisation of non-specific treatment factors.
Another important approach in the quest to improve the results of psychotherapy would be to focus on the possibility that patients with specific characteristics respond differently to specific kinds of treatment; in other words, to learn more about which type of works best for which patients [
22,
23]. If a specific method turns out to work better for specific patients, this would indicate that the use of a particular method probably
does matter.
Personalised therapy using a personalised advantage index
Studies of the use of a PAI (Personalised Advantage Index) found that, if patients receive indicated treatment in accordance with the PAI, there is a relevant and significant difference in treatment effect [
24‐
26]. In this approach, the effect of one kind of psychotherapy for a specific individual by comparison with another kind of psychotherapy is predicted on the basis of an algorithm that is developed using advanced statistical methods [
25,
27]. However, the utility of algorithms in clinical practice has not been established because the models that predict the best possible treatment method for individuals in a specific sample have not yet been validated. A PAI can be built only by using data from an RCT comparing two active treatments [
28,
29]. CBT and SPSP are good candidates for a personalised predictive approach of this kind because they differ in several ways. They are located at either end of a spectrum with respect to the focus of therapy (symptom-centred to more person-centred). The stated mechanisms differ (for example, CBT is thought to change maladaptive cognitions and behaviours, whereas SPSP is thought to be more person-centred), as does the level of structure that is applied during the sessions (CBT is structured and SPSP is not). Lastly, only CBT uses homework assignments.
The second major objective of the study presented here is therefore to build a model that may help predict the optimal type of treatment for a specific individual.
Non-response and therapy switch
At present, despite the prospect of improving treatment outcome if valid prediction models for personalised treatment selection become available, clinical practitioners are still faced with the difficult problem of patients not responding to treatment. Nevertheless, mainly on the basis of expert opinion rather than evidence, several guidelines for the treatment of depression provide recommendations for how to proceed in cases of non-response. For instance, in the United States, the advice of the American Psychiatric Association in its practice guideline for the treatment of patients with major depressive Disorder [
30] is to intensify psychotherapy; to consider a switch to another psychotherapy; or switching to or adding antidepressant medication. The recommendations in the Dutch [
31] guidelines are confined to switching to another kind of psychotherapy or antidepressant medication. The United Kingdom guidelines of the National Institute for Health Care and Excellence advise combining treatments (antidepressant medication and CBT) after initial non-response [
32]. One recent study does indeed conclude that adding antidepressant medication or CBT to initial monotherapy (CBT or medication) leads to an increased response rate [
33]. However, there is no research that supports the advice to switch to another type of psychotherapy when the patient prefers continuing psychotherapy [
34]. A switch of therapy may require a switch of therapist because the initial therapist does not have the necessary expertise, but possibly also because a change in clinical stance is necessary, which could be less credible if delivered by the same person [
35]. In line with the claims made by the advocates of the common-factor theory, the switch to another psychotherapist could in itself make a positive treatment response more likely. The patient could, for example, benefit from a change and establish a better working alliance with a new psychotherapist.
To our knowledge, the study presented here is the first to include a second treatment phase for non-responders to psychotherapy for MDD that addresses the possible effect of a switch of type of psychotherapy and therapist. Non-responders to therapy from the first treatment phase are randomised to three different groups (continuing with the same therapist and the same therapy, switching to another therapist for the same therapy, and switching to another psychotherapist and the other therapy). The third and last major objective of this study is to determine whether a change of therapist or a change of therapist and treatment method are effective strategies to deal with non-response. If the second of these two different strategies proves to be most effective, this could also serve as additional evidence for the relevance of the specific method used. At the least, knowledge about the effectiveness of different psychotherapeutic strategies to deal with initial treatment non-response can be expected to help improve treatment outcome. Even if the strategies studied fail to show any effect, then it will at least be clear that it might be best to switch to a different treatment modality altogether.
The role of the working alliance
A minor objective of the D*Phase study will be to focus on the relevance of the therapeutic alliance as a predictor for treatment outcome. The alliance is commonly defined as the emotional bond established in the therapeutic dyad and the agreement between the therapist and patient about the goals of therapy and the steps necessary to achieve them [
36]. The alliance has been repeatedly shown to be both positively and significantly correlated with treatment effect in different treatment methods [
37]. This has often been interpreted as evidence that the therapeutic alliance is an important “common factor”. However, good alliances may just as well be the
result of changes in symptoms (particularly early changes), rather than a cause [
38]. Causal inferences can only be made if the mediator (in this case, the therapeutic alliance)
precedes the treatment effect. There is a scarcity of methodically robust studies that meet this temporality criterion [
39].
This study will therefore focus on a number of questions relating to the effect of the working alliance on treatment effect. To start with, how do symptom changes affect the working alliance? Does early symptom change predict the quality of the working alliance as perceived by the patient? And does the alliance also have an effect on symptom change in itself, as the common factor theory predicts?
In addition, there are other questions concerning the possible influence of the alliance. For example, does the alliance depend on the characteristics of the therapist, the patient or the match between them [
38,
40‐
42]? The present study also represents a unique opportunity to examine whether, in the case of treatment non-response after treatment phase one, the effect of a switch of therapist or the switch of therapist
and treatment method is mediated by a change in the therapeutic alliance. Should this be the case, that finding can serve as evidence that the quality of the therapeutic alliance is a causal factor in achieving good treatment outcomes.
Treatment integrity and allegiance
Finally, in the quest to identify the factors that are the most relevant targets for the purposes of improving the outcomes of psychotherapy, this study will also focus on the concepts of treatment integrity and allegiance since therapist-related factors may also play a role in the treatment effect.
Treatment integrity consists of two factors: the therapist’s ability to apply therapy-specific techniques as intended (quality or competence) and the extent to which these techniques and methods are applied (quantity or adherence) [
43‐
45]. A recent meta-analysis [
46] of the influence of adherence and competence on treatment outcome across a wide range of mental disorders shows that adherence and competence generally have no significant effect. However, studies that specifically focus on the treatment of depression have found that the competence of the practitioner had a modest effect and reported a trend towards a positive effect of a larger degree of adherence.
The concept of
allegiance is usually used in a scientific context, where it relates to the loyalty of a principal investigator to a specific treatment method [
47]. In the current study, we will focus on the concept of allegiance as it relates to the
therapist and its possible effect on the outcome of psychotherapy
. There is only limited amount of empirical research into the effects of researcher allegiance on the outcome of RCTs, but meta-analyses indicate that allegiance has a clear effect on study outcomes [
48]. It is plausible that, if allegiance affects the outcome of an investigation, similar effects will also be found in the treatment room [
49]. The sparse experimental research that has been done may indicate that a therapist delivers treatment better when he or she has confidence in the particular type of therapy and when he or she is in favour of the underlying principles [
50]. If this hypothesis is correct, it would certainly be worthwhile to investigate whether optimal matching between the indicated therapy and the practitioner who implements it could improve treatment results [
51].
Discussion
The overarching purpose of the study described in this protocol is to acquire knowledge that can help increase the effectiveness of psychotherapy for depression in clinical practice. We aim to do this by addressing a number of research questions in one study with a comprehensive design. Some of these questions have never been the direct focus of a research project. By enrolling depressed outpatients who are randomised to two active treatment conditions (CBT and SPSP), we start the study with a non-inferiority design. This design enables us to address multiple issues. We are able to explore whether SPSP is non inferior to CBT, which extends the evidence base of psychodynamic psychotherapy for depression, which is frequently applied but still relatively understudied. We hope the study will also reveal prescriptive and prognostic factors for treatment response. We will try to build and validate a model which enables us to predict which treatment works best for whom, in other words to personalise treatment. Developing personalised treatment selection strategies is a very promising way to increase the effectiveness of psychotherapy. In the D*Phase study, prediction based on PAI in the first phase of the study can be validated on the basis of the treatment results in phase two of the study.
A novel aspect of the design is found in the second part of the study, which focuses on patients who do not benefit (or who do not benefit enough) from psychotherapy. This is often the case in clinical practice. These patients are randomised again into three groups (continuing the initial treatment with the same therapist, continuing the initial treatment with another therapist and continuing the other type of treatment with another therapist). This provides us with opportunities to find evidence for the expert-based advice given in practice guidelines for the treatment of MDD. The guidelines tell us to switch to another kind of therapy when not enough progress is made but there is no empirical evidence for this advice. Furthermore, we explore what happens when patients switch to another therapist. This is particularly interesting given the assumption made by some researchers that the effects of psychotherapy largely depend on common factors.
By incorporating questionnaires on the working alliance and therapist characteristics that are completed at different points in time, we hope to gain insight into the influence of the working alliance on the effect of psychotherapy. The repeated measures design will provide an opportunity to evaluate whether symptom change occurs in advance of, and is related to, the working alliance. This is one of only a few studies this temporal effect. We also hope to shed light on whether, and which, characteristics of the therapist have an effect on the working alliance and we will address the topic of the relevance of therapist allegiance, with the treatment methods studied, as a predictor of treatment outcome.
The current study is distinctive in the sense that it provides an opportunity to explore the influence of an important common factor (the alliance) and also therapy-specific factors (treatment differences and treatment integrity) on treatment effect in a single, large cohort. Treatment integrity is adequately measured for both conditions. The study also provides several openings for post-hoc analysis. For example, if we find that the therapy and/or therapist switch has an effect, we will be in a position to determine which factors contribute to a possible effect of the switch. If there should indeed be a relationship between competence/adherence and treatment effect, this could be a cautious indication that method-specific interventions do matter in psychotherapy. Of course, findings will only remain global indications and there is no way to measure all possible factors that could play a role.
We therefore intend to implement a pragmatic, methodically strong, study in which the naturalistic setting will contribute to the generalisability of the results to everyday practice. This is a challenging undertaking and it also has certain limitations. Due to the naturalistic setting, for instance, it is not possible to control completely for the use of medication, although the use of medication is monitored and, if necessary, statistically controlled for in the analysis. Different treatment options for depression are available in the treatment centres and patients can choose a different kind of treatment offered by the treatment centre and refused to participate in the research project. This may result in a bias. Secondly, a double-blind design is never possible in a psychotherapy study because the therapist and patient are obviously aware of the kind of treatment that is being administered. However, by digitalising the administration of all the questionnaires that are used, both researchers and therapists are blinded to all patient-reported results and researchers are blinded to the treatment conditions. Thirdly, in spite of the fact that we include several potential predictors, one possible mediator and several moderators, there will certainly be important factors (latent and otherwise) that are not assessed in the current study. Lastly, because the study is primarily powered for the non-inferiority phase, it is expected that phase two of the study only has sufficient power to detect medium to large differences between the groups. Despite this relatively limited power, we still expect this phase of the study will generate meaningful observations on the effects of therapist and treatment switching.
The therapists in the study are thoroughly trained and supervised in the treatment methods concerned in order to ensure that the non-inferiority part of the study will not be biased too much by differences in therapist competence with respect to the different treatment methods. However, most therapists are expected to have more experience with CBT. We will be able to take treatment experience into account as a confounder in our statistical analyses. Furthermore, the measurement of treatment allegiance and treatment integrity will allow for statistical correction for confounding by these factors, at least when it proves to be relevant to do so. There is no instrument to measure treatment integrity for SPSP and so it has to be developed during the course of the study.
Despite these limitations, the design of this study, which compares two different treatments head-to-head, means it will have implications at different levels. In addition to enhancing the evidence base for SPSP, it can teach us more about prescriptive factors for different treatments for MDD and shed light on the role of the therapist in terms of maximising treatment effect by studying the working alliance and the effect of treatment integrity and allegiance. On top of that, this study is the first to manipulate the therapist as a factor through a switch of psychotherapist when therapy does not have the intended effect on depressive symptoms. In conclusion, we hope to provide a large number of new insights that will help to increase the effectiveness of psychotherapy and contribute to the improvement of health care for adults suffering from MDD.
Acknowledgements
Many thanks go to the patients, therapists and also the experts that participated in the development of the adherence scale for SPSP and led the supervision meetings for the therapists (M. Hendriksen, T. Prinsen, R. Kortrijk) and to the teams at Dimence who made the research possible. Many thanks also to: E. van Ankum, A. Schipper, P. van Pelt, S. Stuiver, C. Janssen, M. Vroegop, E. Derksen, A. Shapouri-Moghammadan, K. Blom, J. Keizer, V. Bosch and S. van Bennekom for assisting in the execution of the research; to M. Stender and A. Karst, L. Christhenhusz, the Dimence Group and its research committee for making this study possible and to A. Kaal for his support in data management.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.