Statistics from Altmetric.com
Correspondence to: Jesse Klein, Department of Psychiatry and Behavioural Sciences, Northwestern University, Feinberg School of Medicine, 710 North Lake Shore Drive, Abott Hall, Suite 1205, Chicago, IL 60611, USA;
How effective is cognitive behavioural therapy (CBT) for adolescent depression and what factors explain the observed changes in meta-analytic effect sizes over time?
Effectiveness of CBT; differences in estimates of efficacy of cognitive behavioural therapy for adolescents; moderator variables (treatment duration, nature of sample, type of control group, setting, methodological rigor, therapist vocation, severity of depression at baseline).
Systematic review with meta-analysis.
Medical and psychological databases (PsycINFO and MEDLINE) were searched from January 1980 to September 2006. A hand search of reference lists of studies of CBT was also carried out.
Study selection and analysis:
Published, peer-reviewed randomised controlled trials (RCTs) of CBT in people aged 12–18 years with depression (DSM-III or later, Researcher Diagnostic Criteria, Bellevue Index of Depression Criteria). RCTs had to compare CBT with a control group (for example, waiting list control) or an alternative psychotherapy group (for example, non-directive supportive therapy). Effect sizes were calculated for each study (by dividing the post-therapy difference between depression score means in CBT and control groups divided by their pooled standard deviation, corrected for small sample bias where necessary) and pooled through a random effects meta-analytic model. Cumulative meta-analyses were conducted to evaluate changes in effect of CBT over time. To examine effects of different variables on modification of effect sizes, ANOVA was conducted for the following coded variables: treatment duration (more or less than 867 treatment minutes), sample size, type of control group (active or inactive), setting (clinical or non-clinical), methodological rigour (fulfilling more or less than 17 of 22 CONSORT criteria), vocation of the therapist (clinicians or research assistants/graduate students) and severity of depression.
Eleven randomised studies met criteria for inclusion. Overall, CBT reduced symptoms of depression in adolescents at the end of treatment (mean weighted effect size of CBT: 0.53, 95% CI 0.24 to 0.82; random effects analysis). There was significant heterogeneity between studies. At follow-up CBT also reduced symptoms (mean weighted effect size (0.59, 95% CI 0.14 to 1.05; random effects analysis). Cumulative meta-analysis demonstrated a steady decrease in effect size over time and narrowing of the confidence intervals. Factors contributing to this decline were: improvements in study quality (that is, studies fulfilling more CONSORT criteria had smaller effect sizes; studies using intention to treat analyses had smaller effect sizes), use of active comparator (that is, studies using active comparators had smaller effect sizes), and treatment in clinical settings (that is, studies in clinical settings had smaller effect sizes) (see online table).
Pooling of studies to date suggests that CBT is an effective treatment for adolescents with depression. Methodological differences between early and more recent investigations may be responsible for differences in estimates over time of efficacy of CBT for depressed adolescents.
Klein JB, Jacobs RH, Reinecke MA. Cognitive-behavioral therapy for adolescent depression: a meta-analytic investigation of changes in effect-size estimates. J Am Acad Child Adolesc Psychiatry 2007;46:1403–13.
Cognitive behavioural therapy (CBT) is the most investigated psychosocial or pharmacological intervention for adolescent depression. Early randomised clinical trials (RCTs) suggested that CBT was highly effective for depressed youths, with some of largest reported effect sizes in the adolescent mental health literature. However, recent studies have shown much smaller effects, with some reporting almost no benefit of CBT over control groups. Klein and colleagues probe this apparent drop in effectiveness by comparing the samples, settings, and methods of early and recent RCTs.
Overall, the Klein meta-analysis concludes that the more stringent methods of recent RCTs explain much of the attenuation in CBT treatment effects over the last 10 years. Unfortunately, too few studies were available to tease apart which methodological features were most critical in reducing treatment effects. Examining specific RCTs, there is reason to suspect, however, that sample characteristics may play an important role. For example, within a major CBT RCT, adolescents referred to the trial from clinical sources had worse outcomes than youths ascertained by advertisement, and this difference in referral source was associated with higher levels of hopelessness for depressed teens in clinical settings.1 In the much publicised TADS study, adolescents whose depressions were less severe, chronic, and suicidal were more responsive to treatment, across all treatment types.2 Although not assessed directly by Klein, youth and familial comorbidity also may prove to be an important predictor of CBT effectiveness. Two of the most recent studies in the Klein meta-analysis, with modest effect sizes, included youths with comorbid Conduct Disorder and adolescent offspring of depressed parents. Evidence on the robustness of CBT effects to comorbidity has been mixed.1, 2
Taken together, these results suggest that CBT is likely an effective treatment with mild to moderate cases of “straightforward” adolescent depression. The effectiveness of CBT as a stand-alone treatment in severe, clinically-complicated samples is less clear, although a recent study suggests that CBT may be an effective addition to psychotropic medication for SSRI-resistant depressed adolescents.3 Additional research on the predictors and moderators of CBT effects is much-needed to inform real-world treatment planning for depressed teens and best match interventions to youths and families.
web only appendices 11/3/76
Continuous values were converted into dichotomous outcomes using a median split technique (that is, splitting those above the median into one variable code and those below it into another). Median splits have been criticised for increasing the probability of type I and type II errors. It is unclear what effect this technique would have on the results here.
Files in this Data Supplement:
Source of funding: Fellowship from NIMH to one author.
▸ Additional notes, a table and references are published online only at http://ebmh.bmj.com/content/vol11/issue3
Competing interests: None.
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.