How might quality of life (QoL) be affected by cognitive behavioural therapy (CBT) specifically designed for people with eating disorders? At a general level, we might expect that if a therapy is working well for someone, then naturally their QoL should improve. Without expressing it directly, the authors on a recent meta-analysis (Linardon and Brennan, 2017) take the hypothesis that if a therapy has been shown to be one of the most effective in terms of eating disorder symptoms, then it should also be more effective than other conditions in its effect on QoL.
However, QoL is a complicated concept. It consists of a number of elements, encompassing physical, psychological and social states and both the perceived and objective levels or experiences of these states. Furthermore, QoL is perhaps particularly difficult to measure and understand for people with eating disorders. The egosyntonic aspects of eating disorders may influence individual reports of QoL, as well as the desired states people with eating disorders wish to attain.
The introduction section of this paper (Linardon and Brennan, 2017) is quite direct and concise; whilst likely a stipulation of the journal and beyond the control of the authors, it is a shame that there was not more of a discussion about the value of looking at how therapy influences QoL. Taking a QoL perspective is quite a positive and holistic orientation, as it may be that benefits are experienced in QoL even when therapy outcomes are not reached. These sorts of findings are important for those considering therapy and for therapists themselves. There is the potential to demonstrate how psychological therapy assists a whole person, not just a set of symptoms. Perhaps this is a discussion for another paper or line of research.
In this meta-analysis, 34 papers were identified for inclusion following initial searches on PsychInfo and Medline. These were primary studies, published in English in peer-reviewed journals, of eating disorder specific CBT being undertaken by adults with a range of eating disorders. In these studies, one or more measures of QoL were implemented pre- and post- (and/or at follow-up) treatment. The QoL measures could be subjective general QoL measures or health-related measures (HRQoL) either specific to eating disorders or related to overall health.
The studies included were either RCTs (23 of the studies) or prospective controlled and uncontrolled designs, and the authors discuss how they managed data being drawn from different types of studies. For example, the authors used an aggregation of effect sizes in studies using more than one QoL measure, and they prioritised intent-to-treat (ITT) data over completer data.
The authors used a quality assessment tool I have not come across before (the Quality Assessment Tool for Quantitative Studies, provided by the Effective Public Health Practice Project), which allows papers to be rated as having strong, moderate or weak methodological quality.
As well as looking at the overall effect of CBT on QoL for people with eating disorders, the authors analysed a number of potential moderator variables. These included the type of eating disorder, the type of CBT, the modality of CBT, the quality of the study, whether the data was from ITT or completer participants, whether the study was controlled, and whether the comparator condition was an active treatment or a wait-list.
One paper was not suitable for quantitative analysis and so 33 were included. The authors provide a table listing the key design and variable features and the quality of each study.
For subjective QoL, Hedges’ g=0.50 (p<0.001 for all results unless stated otherwise) and for HRQoL, g=0.55. Within the measures of HRQoL, there were some distinctive differences in effect size, most notably within the eating disorder specific HRQoL measures. For example, the Clinical Impairment Assessment had g=0.80 whilst the Impact of Weight on Quality of Life Scale had g=0.34. Of course, these differences may be due to a number of research factors as well as the difference in measurement, some of which can be hypothesised by looking at the authors’ tables of the features of each study included. It is not clear why the measurement of QoL was not considered as a moderator. How much variation in effect size is related to the different measurements themselves is perhaps worthy of future consideration.
Subjective QoL improvements were significantly larger for participants receiving individual CBT as opposed to group or self-help CBT. For HRQoL, the effect sizes were larger for transdiagnostic samples, and if the therapy was based on the cognitive maintenance model.
For pre-follow-up studies, a larger effect size was found for subjective QoL (g=0.81) but not HRQoL (g=0.52) relative to pre-post designs. Both of these Hedges’ g values are moderate to strong, which suggests, in answer to one of the initial questions in the study, that CBT for eating disorders can have some lasting effect on QoL.
In between-groups studies at post treatment, subjective QoL had a significant g=0.39, but the HRQoL effect size was unreliable owing to the fail-safe number of studies being below threshold. The fail-safe threshold was not met for follow-up studies with a between-group design. I was slightly frustrated to note that the authors describe CBT as effective over inactive conditions in the abstract, without mentioning the failure to meet a fail-safe number of studies to minimise the risk of publication bias.
In their discussion, the authors compare QoL effect sizes for pre-post studies with QoL effect sizes found for studies using CBT to treat bipolar and anxiety disorders, and find these to be similar. For the controlled studies, the QoL effect sizes were smaller than those in studies looking at individuals with depression and with anxiety.
Of the moderating effects, the authors suggest that the larger effect sizes linked to individual therapy may be a dosage effect, as the mean number of sessions was 19.40, compared with 13.87 for group CBT and 8.75 for guided self-help. They comment on the greater effect sizes associated with Fairburn et al’s CBT over other CBT models as being linked to the rapidity of the approach. What is interesting is that both of these moderator effects disappear at follow-up, and the authors consider catch-up effects happening for the other conditions, or problems in the power levels of studies.
In terms of moderators, completer effect sizes were consistently larger than ITT, which seems a logical finding. The authors consider that the larger effect sizes for transdiagnostic samples make sense when one thinks of particular QoL and treatment difficulties for certain types of eating disorder. For example, they suggest that individuals with Anorexia Nervosa may have QoL scores that reflect a degree of egosyntonicity. I am not clear why these effects would disappear simply by grouping together individuals with different diagnoses and it is difficult to unpick this without the details on the samples in each of the transdiagnostic studies.
Aside from exploring the impact of CBT on QoL for adults with eating disorders, the authors had stated their aim was to look at this impact beyond the immediate post-therapy period. It is therefore confusing that this barely features in the discussion, even given the relative lack of studies providing follow-up data. It is perfectly acceptable to discuss findings in the context of their limitations, and to think about what might be being indicated by the existing research. Again, I think a stronger theoretical underpinning to this article would have allowed the authors to do more than simply report the findings.
The authors discuss the inherent issues of unaccountable factors influencing uncontrolled studies, and the difficulties in interpreting their findings given the risks of publication effects. They believe that CBT did have an overall effect on QoL relative to both inactive and active controls, but that there was not enough data to look at what aspects of QoL showed the most and the least change, and that there were not enough RCTs with active controls to properly investigate the impact of CBT over other approaches.
This paper has the potential to have a larger influence on theory development and research design than it would seem at first glance. There is no conclusion section to the paper and little extension into the wider implications of the work. Despite this, the information provided in the paper and the clarity of the research process means that the reader can feel safe to do some further hypothesising of their own, be that around the measurement issues in and between QoL instruments, or the potential for longer-term QoL gains to be maintained following CBT.
The semantics of eating disorder journeys is poorly defined. The words ‘recovery’, ‘relapse’ and ‘remission’ are often used, but there is no consensus on what these terms mean or how they might be established or measured. Khalsa et al (2017) have conducted a review of papers covering these concepts in the context of anorexia nervosa and have suggested a set of criteria that other practitioners can consider. Their guidelines address some of the problems in the literature they reviewed, such as using only BMI or only self-report to confirm recovery or relapse. The guidelines proposed involve a combination of BMI, behaviours, symptoms and the time duration of a given constellation of these factors to indicate recovery, relapse and remission. The paper is well worth reading for practitioners and researchers alike and will hopefully lead to more consistent and holistic criteria, in keeping with other mental health difficulties.
Linardon J, Brennan L. (2017) The effects of cognitive-behavioral therapy for eating disorders on quality of life: A meta-analysis. Int J Eat Disord. 2017 Jul;50(7):715-730. doi: 10.1002/eat.22719. Epub 2017 Apr 21. [PubMed abstract]
Khalsa SS, Portnoff LC, McCurdy-McKinnon D, Feusner JD. (2017) What happens after treatment? A systematic review of relapse, remission, and recovery in anorexia nervosa. Journal of Eating Disorders 2017 5:20 https://doi.org/10.1186/s40337-017-0145-3