Raising the Bar in the Empirical Investigation of Psychotherapy

If you ever go to a conference in which two types of psychotherapy (e.g., cognitive-behavioral and psychodynamic) are compared for the treatment of patients with a particular disorder, be prepared for the arousal of intense emotions in the room. It is curious that otherwise level-headed individuals can become so passionate under these circumstances. It stimulates the interesting question as to how the field of psychotherapy as a whole can advance beyond passionate allegiances to an empirically informed approach to delivering optimal psychotherapies for individual patients.

In this issue of the Journal, Leichsenring et al. (1) report on the long-term follow-up of patients treated in a previously completed randomized clinical trial (2) in which a psychodynamic therapy and a cognitive-behavioral therapy (CBT) were compared in their impact on patients suffering from social anxiety disorder. The design of this randomized clinical trial is an excellent example of the multiple ways scientific methods can rein in identified forms of enthusiastic experimenter bias.

This psychotherapy study is nested in a larger effort, organized as the Social Phobia Psychotherapy Network (SOPHO-NET), with the aim of investigating the treatment, genetics, neural underpinnings, and health economics of social phobia (3). This multifaceted effort of investigating the pathology and its treatments potentially provides an atmosphere of combining greater in-depth knowledge of the mechanisms of the disorder that can be combined with psychotherapy focus and design.

Both the original randomized controlled trial and the follow-up reported here were conducted across five sites that provide a large number of subjects, enabling sophisticated data analysis. In order to control for investigator allegiance to either of the two psychotherapy methods, experts in both CBT and psychodynamic therapy were located at each of the sites. In addition, an independent Coordination Center for Clinical Trials randomly assigned the patients to one of two treatments and monitored the study.

These levels of structural organization were in addition to the careful selection, randomization, and treatment of a large group of patients. As predetermined by a power analysis, 495 patients met inclusion criteria and were randomly assigned to CBT, psychodynamic therapy, or a waiting list. Fifty cognitive-behavioral therapists and 53 psychodynamic therapists specifically trained, closely supervised, and monitored for treatment fidelity were involved. Patient assessments were multiple during treatment and over the follow-up period.

Specific forms of psychodynamic and cognitive-behavioral treatments, manualized with adherence measures, were utilized. An inspection of the two treatment manuals stimulates the realization that the concepts of psychodynamic and cognitive-behavioral are generalizations that carry different meanings and connotations. For research purposes, the treatments are best compared at the level of targets of intervention and strategies used as specified in the treatment manuals. The Clark and Wells CBT model proposes that individuals with social anxiety selectively attend to internal images of themselves that are more negative and less flattering than they actually appear to others. This conceptualization of the problem has led to treatment strategies involving role-playing, restructuring of negative self-image by video feedback, and practicing of external focus of attention. Leichsenring et al. are the first to utilize Luborsky’s psychodynamic approach to core conflictual relationship themes with socially anxious individuals. Core conflictural themes, such as “others will humiliate me,” and “I am afraid of exposing myself,” are examples of the type of themes with these patients. In the specific dynamic treatment used here, themes are explicated and examined in current relationships, including that with the therapist. Thus, the two treatments share common strategies, such as a clear rationale for the pathology combined with a structured approach with encouragement for change, but use different strategies and techniques to achieve symptom reduction.

The follow-up study reported in this issue targets an important gap in the empirical investigation of the treatment of social anxiety disorder, as very few previous treatment studies included follow-up periods longer than 6 months. Follow-up data provide the advantage of putting any treatment differences at the end of treatment into the larger context of potential maintenance or loss of treatment gains and assessment of possible continued improvement. Thus, a comparison of the results at the end of an average of 26 sessions of treatment and upon follow-up provides a view of maintenance of gains and enhanced improvement. Response rates at the end of treatment were 63% (CBT) and 58% (psychodynamic therapy) and 69% for both treatments at 24 months. Remission rates were lower for both treatments, 38% (CBT) and 28% (psychodynamic therapy), and virtually the same at 24 months posttreatment (CBT, 39%; psychodynamic therapy, 38%).

It is not uncommon to find minimal or no differences between two psychotherapeutic approaches to a particular disorder as attested by meta-analysis, especially when the studies are well designed (4). The randomized clinical trial is seen as the gold standard against which any new treatment must show its capabilities. In the randomized controlled trial and follow-up by Leichsenring et al., a new adaptation of Luborsky’s expressive-supportive psychodynamic therapy proved as effective as a more established and creative CBT treatment. Of course, this one study does not resolve the intense debate about the merits of cognitive-behavioral and psychodynamic treatments for this disorder and others. The cognitive-behavioral method of Clark and Wells is seen as innovative and with some proven track record (5), but the dynamic treatment used here can be criticized as not getting to the developmental roots of the dysfunction (6).

More importantly, as many have noted, there are limitations to the standard randomized controlled trial, as these results often lead to no further understanding of mechanisms of change, i.e., if the treatment worked, how did it achieve this? This limitation of the randomized controlled trial design as used in the Leichsenring et al. study is evident; from this set of primary analyses of this landmark study, we are still no closer to understanding mechanisms of change in treatments for social anxiety disorder. The study as designed was a first-level randomized controlled trial, and future studies will bear the burden of examining moderators and mechanisms of change.

So how do the practitioner and the investigator proceed, given these tantalizing yet opaque results? The practitioner 1) can conclude that for social anxiety disorder, either cognitive-behavioral or psychodynamic treatments are equally effective and 2) can opt for the one closest to his or her preferences.

The investigators in this case, Leichsenring et al., emphasize not only that the present study and follow-up yield small differences in outcome between CBT and psychodynamic therapy in the treatment of social anxiety disorder but, most importantly, that the rates of remission and response point to the need for improvement in both treatments. The next step for Leichsenring et al. is the articulation of a unified protocol for the psychodynamic treatment of a range of anxiety conditions (thus, transdiagnostic) by using treatment modules from a number of psychodynamic treatments (thus, unified) (7). This is a positive and potentially fruitful direction, and their next randomized controlled trial may provide needed data on the improvement, given the enhancement of the dynamic treatment.

This choice of direction by Leichsenring et al. is reflective of several developmental currents in the field. There is growing recognition that similar domains of dysfunction are operating across multiple diagnostic categories. For example, overlap in domains of dysfunction is shared by social anxiety disorder, generalized anxiety disorder, panic disorder, and agoraphobia. The so-called transdiagnostic approach (8) utilizes modules of treatment to target core processes across disorders with similar dysfunctions. The transdiagnostic approach to treatment also brings to mind the research domain criteria initiative of the National Institute of Mental Health (9), which has the aim of focusing research not on particular DSM diagnostic categories or entities but rather on domains of dysfunction that extend across the categorical disorders. However, there are urgent and intense calls for psychotherapy research to approach the central issue of causality and the mechanisms of change that the treatment in question targets. It is only with the specification of mediators of change, whether psychological (10) or neurobiological (9) in nature, will the field move forward. With their focus on the psychology and neurobiological aspects of social anxiety disorder, Leichsenring et al. are poised to approach these central issues.