Statistics from Altmetric.com
Systematic review with meta-analysis.
PsycINFO and MEDLINE (1980–2000); hand searches of 19 key journals and reference lists of published meta-analyses and reviews.
Study selection and analysis:
Randomised controlled trials, using validated outcome measures, comparing mainly face-to-face psychosocial treatments for bulimia with either a control condition or an alternative treatment were included. Exclusions: naturalistic studies, studies in uncommon subtypes of bulimia, not published in English, or fewer than 10 participants. Data were extracted on number of participants, losses to follow up, effect size for binge eating and purging (Cohen’s d; treatment-control and pre-post), percentage recovery (defined as complete recovery or recovery from purging), mean post-treatment symptoms (mean weekly binge eating and purging episodes), and percentage of participants seeking further treatment during follow up.
Recovery from binge eating and purging.
Twenty six trials met inclusion criteria. The trials covered 51 treatments: 25 cognitive behavioural (CBT), 11 behavioural (BT), and 15 other therapeutic approaches. Treatments used individual therapy (30 treatments), group therapy (19 treatments), or a combination of these (2 treatments). Mean treatment length was 16.6 sessions (range 6 to 52 sessions). On average 82.3% of participants completed the treatment (46 treatments). The overall recovery rate was 33% (ITT analysis); recovery rates were higher with individual therapies compared with group therapy (p = 0.004). The overall mean effect size for treatment v control was 0.88 (see table⇓).
Psychotherapy improves bingeing and purging in people with bulimia nervosa. Individual psychotherapy is more successful than group psychotherapy (using CBT or BT).
Authors note that on average 40.1% of people screened were excluded from the trials, this may affect the generalisability of results to people treated for bulimia nervosa in the community. This meta-analysis included heterogeneous psychotherapeutic interventions, which varied in theoretical basis and methods of application, making it difficult to draw conclusions about specific interventions.
Thompson-Brenner and colleagues’ meta-analysis presents several novel findings. The superiority of individual compared with group psychotherapy is one of their most consistent results, leading the authors to conclude, “investigators and clinicians should focus on individual rather than group CBT” (p 278). However, this conclusion may not result in dramatic changes in current research or practice, as most investigators and clinicians have already made this transition. The appendix of studies included in their meta-analysis reflects how group therapy appears to have fallen out of favour for the treatment of bulimia nervosa. Among studies published in the last decade (1994–2003), two included group interventions while the remaining seven examined only individual interventions. Indeed, the average publication date for studies of group and individual interventions were 1989 and 1994, respectively (t (21) = 2.3, p = 0.03).
The positive association between publication year and CBT’s effect size raises the question of whether investigators should focus less on group CBT because it is less effective, or if group CBT appears to be less effective because investigators have focused less on it in recent years. To examine this question, one can re-assess the difference between individual and group therapies from studies published before 1994. In contrast to the authors’ original results, such analyses show no significant differences between group versus individual treatments on post-treatment binge (F (1,18) = 0.25, p = 0.63) or purge frequency (F (1,25) = 0.62, p = 0.44) (pre-treatment binge and purge frequencies were entered as covariates to control for initial differences between group and individual treatments). The superiority of individual over group therapies may have evolved from pursuing and improving individual therapies while abandoning group interventions before any true differences in efficacy emerged. Thus, rather than anticipating a significant impact of meta-analytic findings on clinical and research practices, it seems that clinical and research practices may have significantly impacted meta-analytic findings.
For correspondence: H Thompson-Brenner, Center for Anxiety and Related Disorders, Department of Psychology, Boston University, Beacon Street, Boston, Massachusetts, USA;
Source of funding: The Glass Foundation.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.