Have the potential benefits of CBT for severe mental disorders been undersold?

Thase et al's useful review and update on cognitive behavior therapy (CBT) for severe mental disorders (1) highlights the recurrent debate about the magnitude of the impact of CBT on persons with psychosis, including the question raised by some as to whether the effectiveness of CBT for this population has been “oversold” (2,3). We would like to take this opportunity to adopt the opposite position, and to suggest that rather than the benefits of CBT being oversold, the potential benefits have actually been undersold. We argue our point by discussing three issues related to the evaluation and application of CBT for persons with severe mental disorders, including inattention in meta-analyses to critical study variables, over-emphasis on comparing CBT to so-called non-specific therapy controls, and the relative lack of use of CBT in persons with psychosis to address the most conventional treatment targets for which CBT has been established in the general population.

The conclusions drawn by meta-analyses are only as accurate as the variables summarized from the original studies. Researchers who conduct meta-analyses go to great lengths to measure a broad range of study characteristics, such as using standardized approaches to quantify methodological rigor, publication bias, and the outcomes themselves, and this has been true in work evaluating the outcomes from CBT for psychosis (2,4,5). However, despite this laudable attention to capturing study details, meta-analyses of research on CBT typically ignore the unique outcomes targeted by specific studies, and combine all available studies when evaluating the impact of CBT on symptom and functional domains (2). Because of the breadth of potential targets of CBT for psychosis, these procedures may inadvertently lead to drawing inaccurate conclusions.

CBT comprises a broad range of treatment elements that can be adapted and applied to a wide variety of symptoms and impairments, and the specific treatment target in any one study has important implications for interpreting the results of meta-analyses which include that study. Studies may target positive symptoms, negative symptoms, relapse, and/or functional impairments. Combining studies that target different outcome domains is like “comparing apples to oranges”, and can lead to an underestimation of the effects of CBT when the results of studies that did not target a specific domain are combined with those that did. For example, Sensky et al (6) evaluated the impact of CBT for psychosis on stable outpatients with persistent positive psychotic symptoms, Garety et al (7) focused on relapse prevention and psychotic symptoms in individuals who had recently experienced a relapse of psychosis, and Granholm et al (8) targeted defeatist attitudes and functional impairment in persons with schizophrenia. However, the results of all three studies were pooled in a recent meta-analysis examining the effects of CBT on psychotic symptoms (2), which likely obscured the results.

In a related fashion, inattention to critical population characteristics or contextual factors related to with whom and where the study was conducted can lead to results that are biased against CBT. For example, Lewis and colleagues (9,10) evaluated the impact of adding CBT for psychosis to inpatient treatment for people recently hospitalized for a first or second episode of psychosis. Psychosis severity decreased dramatically for the CBT, the supportive therapy, and routine treatment groups from baseline to post-treatment and follow-up, with the CBT group showing slightly more rapid improvement, but no differences between the groups at follow-up. The study appears to be a failure of CBT. However, what does it teach us about the effects of CBT in the sizable population of people with schizophrenia-spectrum disorders who experience persistent psychotic symptoms and are the typical recipient of the treatment? Very little. And yet in a recent meta-analysis (2) this study was included along with other studies of CBT for psychosis conducted in very different settings, such as studies focusing on stabilized outpatients with psychotic symptoms.

Meta-analyses are a powerful tool for evaluating the impact of an intervention by combining the results of multiple studies. However, the choice of which studies to include is not a trivial one, and needs to be made with an understanding of the nature of the intervention and target population. In the case of CBT, the penchant for meta-analyses to indiscriminately combine the results of studies targeting different outcomes in different populations or contexts has likely led to an underestimation of the true impact of CBT on selected targeted outcomes.

In most treatment systems for people with a serious mental disorder, resources to provide psychotherapy are limited. The primary question facing clinicians and policy makers is whether adding a particular form of treatment will significantly improve symptoms or functioning compared to the often-constrained services as usual. This question is particularly apt when discussing the inclusion of psychotherapy, which tends to be labor-intensive and not routinely provided to this population, even when evidence supports it, in contrast to pharmacological treatment, which is often the standard of care even in the absence of compelling evidence to support the specific intervention (e.g., off label use, polypharmacy). Unfortunately, most of the controlled research evaluating the effects of CBT has not been designed to address this most basic question facing clinicians and policy makers: does CBT for psychosis added to customary care confer more benefits than customary care alone?

In the field of mental health, researchers usually think about the utility of interventions in very different ways from clinicians. In their zeal to prove that psychotherapy can be studied just as rigorously as medication, psychotherapy researchers have typically adopted a “gold standard” control group modeled after the placebo used in randomized pharmacological studies. The rationale for using a control psychotherapy intervention is not based on the presumed inertness of the intervention, but rather that it controls for “non-specific treatment factors” common across all psychotherapies. Psychotherapy researchers are under tremendous pressure to adopt research designs that compare a specific psychotherapy model (e.g., CBT) with a potentially active control psychotherapy (e.g., supportive therapy, befriending), rather than treatment as usual, because such designs are considered to be more “rigorous” or elucidating of the mechanisms underlying the experimental treatment by reviewers. However, the answers obtained using this research design typically have limited practical utility to settings where supportive therapy or befriending are not the standard of care. Comparing a psychotherapeutic intervention to another “control” psychotherapy is informative about the relative benefits of one intervention over the other, but not about the absolute benefit of the experimental intervention when added to usual services, the primary question at stake. Research (including meta-analyses) that concludes that CBT offers little more than a control condition such as supportive therapy inadvertently misses the critical issue for most clinicians: is there a strong probability that this intervention will improve outcomes for the individuals with serious psychiatric illnesses with whom I work?

Aside from the fact that studies comparing CBT to a control psychotherapy do not provide a direct test of the added benefit of CBT in a typical clinical setting, there are at least two other problems with such research designs. First, designs that evaluate the relative benefit of CBT vs. another psychotherapy approach provide no reasonable basis for even inferring the degree of improvement provided by CBT over usual care. Unlike medication placebos, the effects of non-specific psychotherapeutic interventions for people with severe mental disorders are not well understood, hence even when CBT outperforms a control psychotherapy, the absolute benefit of CBT remains unclear. Second, there is modest evidence suggesting that supportive psychotherapy may be beneficial for people with schizophrenia (11). To the extent that “non-specific treatment factors” in psychotherapy do contribute to improved outcomes in people with severe psychiatric disorders, research designs that compare CBT with a control psychotherapy will underestimate the effects of CBT over usual care in a typical clinical setting, where even supportive therapy may be a rare commodity.

Although the primary evidence base for CBT lies in decades of research on its effects on depression and anxiety, somewhat surprisingly these two symptom domains have not been the focus of extensive CBT research in people with schizophrenia and other disorders with psychotic features. Depression frequently antedates the onset of schizophrenia, and is one of the most consistent and impairing clinical syndromes in the illness (12). Research trials of CBT for psychosis often target depression related to psychotic symptoms, and demonstrate positive effects in reducing depression. However, very limited research has evaluated the effects of CBT on depression as the primary target symptom in this population (13,14), suggesting that CBT may be underutilized, and its effects under-appreciated in people with psychotic disorders.

Similar to depression, anxiety and anxiety disorders are also common in people with schizophrenia and related disorders, but have not been the focus of extensive CBT research. Some research does suggest that CBT is effective for anxiety disorders such as post-traumatic stress disorder (15) and social phobia (16) in people with severe psychiatric disorders. If CBT produces similar or even attenuated effects on anxiety in people with schizophrenia and other severe mental disorders as those reported in the general population, it would be an important tool for clinicians to alleviate much of the suffering caused by these disorders.

In conclusion, a growing body of research supports the efficacy of CBT for severe psychiatric disorders, but, as Thase et al note, access to CBT remains limited for most people with these conditions. While there has been an ongoing debate as to the magnitude of the impact of CBT, there are strong reasons to suggest that reviews of the research literature and commonly employed research designs have led to a systematic underestimation of the benefits of adding CBT to usual services. Specifically, meta-analyses that have failed to take into account the treatment targets, study populations, and contexts of different studies of CBT may underestimate treatment effects by combining the results of studies focusing on very different outcomes. Research designs that compare CBT with another active psychotherapeutic intervention in order to control for “non-specific treatment factors” will underestimate the incremental benefit of adding CBT to usual services to the extent that such control treatments produce some clinical benefit. Last, CBT has had only limited application to the problems of depression and anxiety as a primary treatment focus for people with schizophrenia and related disorders, despite the preponderance of evidence supporting the effects of CBT for these symptoms in the general population. This suggests that CBT may have untapped potential for addressing these problems in people with severe mental disorders. The confluence of these factors, combined with the lack of access to CBT for most people with severe mental disorders, suggest that the true benefits of CBT have been undersold, not oversold, to clinicians, researchers, policy makers, consumers of mental health services, their families, and the public at large.