Mindfulness‐based cognitive therapy and depression relapse—evaluating evidence through a meta‐analytic lens may indicate myopia

For those affected by Major Depressive Disorder (MDD) relapse remains a major problem, with 35% of people with a major depressive episode experiencing at least one relapse.1 Interventions to prevent relapse include maintenance antidepressants, as well as a number of psychosocial interventions, amongst them Cognitive Behavioural Therapy (CBT), Interpersonal therapy (IPT), Behavioural Activation and Mindfulness-Based Cognitive Therapy (MBCBT). MBCBT is a relatively recently developed psychotherapy, consisting of manualised group skills training, based on the principles of CBT and mindfulness-based stress reduction. It was developed specifically for relapse prevention in depression and has been investigated in a number of randomised controlled trials (RCTs). It is recommended as a maintenance treatment for depression in a number of guidelines, amongst them the UK National Institute for Health and Care Excellence (NICE),2 Canadian Network for Mood and Anxiety Treatments (CANMAT),3 and Royal Australian and New Zealand Clinical Practice Guidelines (RANZP).4 In this issue of Acta Psychiatrica Scandinavica, McCartney and colleagues present the results of a pairwise and network meta-analysis of mindfulness-based cognitive therapy for prevention of relapse of depression and time to relapse.5 Any attempt at synthesis of the literature in this field is to be welcomed and will likely have important clinical ramifications. By conducting a network meta-analysis, the authors also attempt to bypass the limitations of including only trials involving direct comparisons, thus enabling further comparisons. This approach may therefore add to existing evidence from pairwise and individual patient data meta-analysis. The headline findings appear to give cause for optimism versus treatment-as-usual (TAU) MBCBT shows a statistically significant difference in preventing depression relapse and, compared to TAU and placebo, an increase in time to relapse (though only one trial contributed to the placebo comparison). This is in keeping with findings of an individual patient data meta-analysis, which showed similar results for reates of relapse.6 Therefore, on the face of it, this meta-analysis provides support for guideline recommendations, and possibly the wider use of this intervention. However, meta-analyses (both pairwise and network) are only as strong as studies included and close examination of evidence contributing to this meta-analysis reveals pertinent limitations in trial methodology. Cuijpers and Cristea neatly outline some of these factors7 which include, though are not limited to, allegiance bias, expectancy effects, choice of comparator (active controls versus waiting list), small sample size and selective reporting. Any novel therapy examined by its proponents is at risk of allegiance bias—for some reason or another, effect sizes are higher in trials conducted by those who have an interest in the novel technique and may have more expertise in delivering the intervention. Examining studies included here, a number of larger trials were performed by original proponents of MBCBT. These trials were generally positive and contributed significantly to the overall results of the meta-analysis. It should be acknowledged that at the time of the original Teasdale trial (in 2000), there was little evidence for psychotherapies in preventing depression, and therefore, this was a novel trial. Nonetheless, readers should be aware that aspects of methodology were not as stringent as one would expect in trials today (eg initial stratification of the sample for randomisation by 2 or more episodes of depression, intention to treat analysis of people with three or more episodes, as opposed to people with two or more episodes,8 Non-significant findings for people with 2 or more episodes may have been the result of decreased statistical power, though this is unclear.9) The follow-up study with Teasdale as main author10 demonstrated a significant difference for people with three or more episodes, though not for two prior episodes. The treatment-as-usual group experienced a 100% relapse rate. In terms of comparators, it should come as little surprise that in the meta-analysis, this intervention showed a benefit compared to TAU, though not in comparison to other active interventions such as minimal antidepressant medication, an active control condition, cognitive psychological education, cognitive therapy and depression relapse active monitoring. 2 of the trials in the TAU group are categorised as ‘waiting list controls’, and whilst this may be considered TAU in primary care, it is worth noting that one of these trials showed the greatest effect size in the meta-analysis. People taking part in the intervention group in novel studies may well have expectancy effects, and therefore, without an adequate placebo control group, it is impossible to tell what component of an intervention is helpful. This could be compounded by possible nocebo effects in the TAU group, more so in a waiting list control group. By including pilot trials/feasibility studies with small sample sizes, the odds of finding an effect are increased, through effects such as outliers. Furthermore, participants in these studies show expectancy effects as noted above, and when larger RCTs are run, effect sizes diminish. Lumping these studies together tends to give an artificially elevated effect size. These concerns have been taken on by those in this field—PREVENT was an RCT which showed no difference between maintenance antidepressant medication (ADM) and MBCBT, and was adequately powered (N = 424). Its primary outcome was time to relapse and the hypothesis was that MBCBT would show an advantage above maintenance antidepressant medication over the treatment period of 24 months. No statistically significant difference was found, and it should be noted that some participants in the MBCBT group (who were to stop ADM) continued to take antidepressants during the trial. Though this is an advantage in terms of pragmatism (real-world processes), it does little to address the original research question. The lack of a placebo group also makes inference difficult.11 This trial does, however, provide evidence for those in primary care with recurrent depression who wish to stop antidepressant medication and may therefore wish for a psychosocial intervention to decrease relapse risk. The above critiques are not limited to trials of MBCT and are well illustrated in trials of psychosocial interventions for other major mental illnesses, such as schizophrenia.12 Perhaps, the most relevant issue is the use of TAU as comparator. A number of RCTs have been conducted where people with schizophrenia or psychosis are offered CBT as an alternative to antipsychotics, utilising TAU comparator groups. These have included CBT versus antipsychotics (TAU) or a combination of both13, 14 (or a combination of CBT plus family therapy versus antipsychotics or both).15 The results of these trials have led some to suggest that ‘patient choice’ should include the use of this intervention in place of antipsychotic medication (cited in the current meta-analysis). All of these trials have been conducted by the same research group (Morrison and colleagues) and have TAU as comparator (with no active psychological placebo). One trial showed a significant difference between CBT and TAU, with 37 people in each arm,13 one an advantage for CBT plus antipsychotics over CBT alone,14 and one (feasibility study) showed no difference between psychosocial interventions and TAU in adolescents with psychosis.15 The only other trial examining this question (CBT as an alternative to antipsychotics) was a recent non-inferiority RCT of CBT case management (CBCM) plus antipsychotic or CBCM plus placebo, in 15–25-year-olds with a variety of psychotic disorders (<40% with schizophrenia spectrum disorder) attending a specialist early intervention service, in Melbourne.16 No difference in the primary outcome of functioning at 6 months was found. Of the 90 participants (constituting around 7% of eligible participants), there were high drop-out rates in both groups (placebo and risperidone/paliperidone), a significant number of those randomised to placebo ended up on antipsychotic medication, and doses of antipsychotic could be considered low (only around 40% taking risperidone ≥ 3 mg). Therefore, even with a more rigorous study design (and comparator), existing evidence for CBT as an alternative to antipsychotics is not very strong. How do we make sense of this literature, and what does it mean for funders, guideline groups, and most importantly patients and their families? Patients, and their families, often welcome psychological therapies and perceive these to have less side effects (though the reporting of side effects in psychotherapy trials is historically poor). However, within the economic constraints of any healthcare system, all interventions should be adequately scrutinised, and this applies to psychotherapy trials. From the findings of the current meta-analysis, it would seem reasonable to assert that TAU comparisons do little to contribute to advancing our knowledge of the effectiveness of a psychosocial intervention, and therefore, it seems difficult to justify the use of TAU in future trials. The authors would like to acknowledge Peter McKenna who commented on an earlier version of this manuscript. AHY reports paid lectures and advisory boards for the following companies with drugs used in affective and related disorders: AstraZeneca (AZ), Eli Lilly, Lundbeck, Sunovion, Servier, LivaNova, and Janssen. No shareholdings in pharmaceutical companies. Lead Investigator for Embolden Study (AZ), BCI Neuroplasticity study and Aripiprazole Mania Study. Investigator initiated studies from AZ, Eli Lilly, Lundbeck, Wyeth, Janssen. SJ has received honorarium for lectures from Sunovian. KCL has received honoraria from Lundbeck for lectures SJ has given. The peer review history for this article is available at https://publons.com/publon/10.1111/acps.13270.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.