Summations
• Evidence of the superiority of combined treatment [antidepressant and cognitive behavioural therapy (CBT)] in comparison with antidepressant alone in both depressive and anxiety disorders has been found after 12-week treatment.
• A higher global functioning [Children's Global Assessment Scale (CGAS)] and a more marked improvement [(Clinical Global Impressions Improvement (CGI-I)] was found in patients treated with the combined treatment for 12 weeks.
• Combined treatment seems to have an unspecific effect on global functioning and general improvement rather than on depressive symptomatology.
Considerations
• The pooling of studies on both major depression and anxiety disorders obviously represents a limitation from the point of view of diagnostic homogeneity.
• The reported results pertain to the acute treatment outcome (12 weeks) and could not be extended to longer follow-up.
• The use of different antidepressant classes and molecules represents a further source of heterogeneity.
Introduction
The research evidence on the effectiveness of antidepressants and their risk-benefit ratio in adolescents is a debated issue (1–6). High caution has been suggested regarding antidepressant administration in this age group (Reference Papanikolaou, Richardson, Pehlivanidis and Papadopoulou-Daifoti2,Reference Masi, Liboni and Brovedani5) and specific indications for psychological or at least combined therapies as first-line treatments in young patients have been reported (Reference Gentile6).
However, according to guidelines, the prescription of antidepressants in adolescents should be considered in the case of moderate-to-severe depression or anxiety which fails to respond to psychological therapy (7–11), always within the context of a comprehensive management of the patient. This raises the issue if antidepressants should be prescribed in combination with psychological therapies or alone.
A few recent studies focused on the comparison between antidepressant alone versus combined therapy (Reference Tandon and Pruett12), filling up the gap of earlier research (Reference van Engeland13). In particular, three multi-centre, controlled, publicly funded clinical trials have been performed on this topic on major depressive disorder (MDD) patients: the Treatment of Adolescent Depression Study (TADS) (Reference March, Silva and Petrycki14), the Treatment of Resistant Depression in Adolescents (TORDIA) (Reference Brent, Emslie and Clarke15), and the Adolescent Depression and Psychotherapy Trial (ADAPT) (Reference Goodyer, Dubicka and Wilkinson16).
The National Institutes of Health-funded TADS study investigated the efficacy of 12-week cognitive behavioural therapy (CBT) in comparison with fluoxetine alone, fluoxetine plus CBT, and placebo in 439 adolescents with moderate-to-severe depression (Reference March, Silva and Petrycki14). According to the reduction in the Children's Depression Rating Scale – Revised (CDRS-R), combined treatment was superior to all other treatments, while according to the rate of Clinical Global Impressions Improvement (CGI-I) responders, fluoxetine alone and combined treatment did not differ but they have been found to be both superior to CBT alone and placebo. In the same line of evidence, the National Institute of Mental Health-funded TORDIA study showed that in 334 depressed adolescents who have failed to respond to a selective serotonin reuptake inhibitor (SSRI) at adequate dose and duration, a switch to a 12-week combined therapy consisting in another antidepressant plus CBT resulted in a higher response than a switch to another antidepressant alone (Reference Brent, Emslie and Clarke15). However, from the National Health Service-funded ADAPT trial emerged a contrasting result: fluoxetine plus routine monitoring has been compared to fluoxetine plus CBT in 208 depressed adolescents who did not respond to a brief psychosocial intervention and the combined treatment was not associated with an improved outcome in comparison with SSRI alone (Reference Goodyer, Dubicka and Wilkinson16).
In another MDD study comparing a treatment as usual (TAU) control condition consisting primarily of SSRI medication and the combined treatment with brief CBT, the combined treatment did not show detectable improvement on the Children's Global Assessment Scale (CGAS) (Reference Clarke, Debar and Lynch17).
Pertaining to anxiety, results run in the same direction of studies on MDD: a randomised controlled trial (RCT) on 488 adolescent patients with a primary diagnosis of separation anxiety disorder, generalised anxiety disorder or social phobia reported that the combination of sertraline and CBT were more effective in comparison with sertraline alone, CBT alone and placebo (Reference Walkup, Albano and Piacentini18).
Recently, an extensive meta-analysis of RCTs focused on newer-generation antidepressants and CBT in adolescent depression has been performed (Reference Dubicka, Elvins, Roberts, Chick, Wilkinson and Goodyer19).
Aims of the study
We performed a new meta-analysis with the purpose of providing a more in depth estimation of combined treatment effect after 12-week treatment (comparison data with the previously published meta-analysis (Reference Dubicka, Elvins, Roberts, Chick, Wilkinson and Goodyer19) are available on request). Moreover, we focused on the comparison between antidepressant alone versus combined treatment with antidepressant plus CBT not only in major depressed but also in anxiety disorder adolescents. In fact, our specific aim was to evaluate antidepressant impact on the global functioning of young patients throughout different diagnoses. However, we further performed sensitivity analyses on diagnostic homogeneous samples.
Methods
Search strategy
An electronic search of the literature was performed to identify studies having as a primary outcome the investigation of the effectiveness of antidepressants alone versus combined treatments in adolescents.
PubMed, PsychINFO and Cochrane databases were searched for articles published until April 2011 using any combination of the terms ‘depressive disorder', ‘antidepressant', ‘combined treatment', ‘CBT' and ‘randomised controlled trials'. Reference lists from identified articles and reviews were used as well to find additional studies to be included.
Study selection
Fifteen reviewers independently screened the yield from the searches to identify potentially relevant studies. Studies were included if they: (a) were written in English language; (b) included adolescent patients affected by a psychiatric disorder diagnosed with the Diagnostic and Statistical Manual of Mental Disorders or the International Classification of Diseases and treated with an antidepressant; (c) showed comparison between antidepressant alone versus combined therapy; (d) include CBT therapy in the combined arm; (e) showed scores of at least one of the most used scales to assess symptom severity or clinical improvement in this population; and (f) showed a follow-up at 12 weeks. Moreover, we decided to perform the analyses on a specific scale in the presence of at least three studies reporting the related scores.
Studies were excluded if: (a) they were performed on overlapping samples and (b) no scores of symptomatological improvement or severity scales were separately given for the two trial arms.
Outcome measures
Primary outcome. The primary outcome was the comparison between antidepressant alone versus combined treatment groups in terms of scores at the CGAS through different diagnoses. Mean scores and standard deviations have been compared.
Secondary outcomes. The secondary outcomes were the comparison between antidepressant alone versus combined treatment groups in terms of scores at the CDRS-R and at the CGI-I. For CDRS-R mean scores and standard deviations have been compared, while for the CGI-I the number of patients reporting a CGI improvement score of 1 (very much improved) or 2 (much improved) has been compared with the total. Moreover, as secondary outcomes we performed some sensitivity analyses to separately consider MDD patients for each scale.
Data extraction
For each study, the following information were extracted: first author, publication year, total sample size and sample sizes for each sub-group, diagnostic status, gender ratio, mean age and age ranges, reported ethnicity, treatment and daily dose, CBT number of sessions and duration, duration of follow-up, scales used for the assessment and main results (Table 1).
Studies included in the meta-analysis
BDI, Beck Depression Inventory; CBCL, Child Behaviour Checklist; CES-D, Center for Epidemiological Studies–Depression Scale; CGI-S, Clinical Global Impressions Severity; HNOS, Health of the Nation Outcome Scales; MFQ, Mood and Feelings Questionnaire; NA, not applicable; NS, not significant; PARS, Pediatric Anxiety Rating Scale; RADS, Reynolds Adolescent Depression Scale; SAS-SR, Social Adjustment Scale-Self Report for Youth; SF-12, Short Form-12; SIQ, Suicidal Ideation Questionnaire; YSR, Youth Self-Report.
Assessment of quality
The quality of the included studies has been assessed using the Jadad Scale (Reference Jadad, Moore and Carroll20). This instrument assesses the quality of (RCTs) allocating a total score comprised between zero (very poor) and five (rigorous). It assessed the randomisation, the blinding and the withdrawal description. Furthermore, potential publication bias was assessed using both funnel plots and Egger test (Reference Egger, Smith and Phillips21).
Data analysis
Data were entered into the Cochrane Collaboration Review Manager software (RevMan version 5.1). Individual and pooled 95% confidence intervals (CIs) were calculated. Heterogeneity between studies was assessed with chi-squared test of fit. The significance of the pooled effect size was determined using a Z-test.
Data were analysed within a random effect framework because of the assumption of the presence of significant between-study heterogeneity. In the first analysis we included the whole studies. In further sensitivity analyses we removed some studies as described to reduce the heterogeneity.
Results
Inclusion/exclusion process and study description
The initial search yielded 2023 articles. Titles and abstracts were carefully revised in order to investigate if they fulfilled inclusion criteria. Of these articles, 1974 have been excluded because they were not performed on adolescent samples, they were not RCTs or they were reviews. The full texts of the 49 remaining studies were obtained and the reviewers independently applied inclusion/exclusion criteria. Forty-four studies were then excluded because they did not show two arms in which antidepressant alone versus combined therapy were tested or they were performed on overlapping samples (see Fig. 1 for the flow diagram).
Flow diagram of the study inclusion/exclusion process.
Five studies met the criteria for inclusion in this meta-analysis. A summary of the included studies is provided in Table 1. Of these, four studies included MDD samples (14–17) and one included anxiety disorder sample (Reference Walkup, Albano and Piacentini18).
The methodological quality of the included studies according to the Jadad Scale (Reference Jadad, Moore and Carroll20) is shown in Table 1. Three of the four included studies were RCTs with a high methodological quality (score of 5). One study (Reference Goodyer, Dubicka and Wilkinson16) could not be evaluated as it was a pragmatic randomised superiority trial.
Primary outcome
We firstly performed an analysis including both MDD and anxiety diagnoses for the CGAS scale. Random effects indicated an association between higher CGAS scores (higher global functioning) in the combined treatment group (Z = 4.16, p < 0.0001) without evidence of between-study heterogeneity (X 2 = 2.68, df = 4, p = 0.61, I 2 = 0%) (Fig. 2). The funnel plot showed a symmetric inverted shape. Egger's test indicated no evidence of publication bias (p = 0.32).
Children's Global Assessment Scale scores in antidepressant versus combined treatment groups.
Removing the anxiety disorder study in a further sensitivity analysis the association remained significant (Z = 3.16, p = 0.002) without heterogeneity (X 2 = 2.13, df = 3, p = 0.55, I 2 = 0%).
Secondary outcomes
Considering the CGI-I scale, significantly higher rate of improved patients has been found in the combined treatment group (Z = 2.02, p = 0.04) (Fig. 3). Nevertheless, as expected, there was evidence of between-study heterogeneity (X 2 = 11.28, df = 3, p = 0.01, I 2 = 73%). The funnel plot showed a symmetric inverted shape. Egger's test indicated no evidence of publication bias (p = 0.88).
Clinical Global Impressions Improvement in antidepressant versus combined treatment groups [patients reporting a CGI improvement score of 1 (very much improved) or 2 (much improved) have been compared with the total].
In a sensitivity analysis, we removed the only study performed on anxiety disorder patients to observe if it consistently affected the heterogeneity. Random effects indicated no more between-study heterogeneity (X 2 = 2.65, df = 2, p = 0.27, I 2 = 25%) with results in the same direction but with a reduced significance level (Z = 1.75, p = 0.08).
Finally, considering the CDRS-R scale, no difference has been found between antidepressant alone and combined treatment (Z = 0.74, p = 0.46), without evidence of between-study heterogeneity (X 2 = 4.18, df = 2, p = 0.12, I 2 = 52%) (Fig. 4). In this case, the funnel plot showed a non-symmetric shape and Egger's test indicated evidence of publication bias (p = 0.04).
Children's Depression Rating Scale – Revised scores in antidepressant versus combined treatment groups.
Discussion
Higher global functioning in patients treated with the combined treatment. The aim of this meta-analysis was to compare the efficacy of antidepressant treatment alone versus the combination of an antidepressant with CBT in adolescent population with different psychiatric diagnoses (MDD and anxiety disorders). Present results are similar to the ones reported in adult samples (Reference Molenaar, Dekker, Van, Hendriksen, Vink and Schoevers22,Reference Sharp, Power, Simpson, Swanson and Anstee23): a mild evidence of the superiority of the combined treatment in both depressive and anxiety disorders has been found after 12-week treatment.
In detail, a higher global functioning (CGAS) was found in patients treated with the combined treatment for 12 weeks. Similarly, a more marked improvement (CGI-I) has been found in the same group. On the contrary, when depressive symptomatology has been considered, no effect of combined treatment was found. These findings substantially confirmed the ones reported by Dubicka end colleagues (Reference Dubicka, Elvins, Roberts, Chick, Wilkinson and Goodyer19), but with a higher significance regarding global functioning (p < 0.0001 considering anxiety disorders and p = 0.002 without considering them).
Consequently, combined treatment seems to have an unspecific effect on global functioning and general improvement rather than on depressive symptomatology. This result could be due to a plausible CBT specific effect on the promotion of coping strategies and healthy behaviours rather than on the correction of cognitive distortions associated with depression in this population. This is in line with evidence of a higher efficacy of CBT in patient subgroups characterised by milder depression (Reference Curry, Rohde and Simons24).
Anxiety disorders seem to have the most favourable outcome, in line with findings from TORDIA trial, in which the number of comorbid anxiety diagnoses was a positive moderator of combined treatment effects (Reference Asarnow, Emslie and Clarke25).
The fact that ADAPT results run in the opposite direction from other studies (Figs 3 and 4) could be interpreted considering that only non-responders to an initial psychotherapy were randomised into this trial and consequently patients less responsive to psychotherapy might have been selected.
Limitations
To the best of our knowledge, this is the first meta-analysis on this topic covering both depressive and anxiety disorders. A strength of this study was the quality assessment of the included trials through a specific scale.
However, several limitations have to be listed. First of all, the pooling of four studies on major depression together with a trial on anxiety disorders obviously represents a limitation from the point of view of diagnostic homogeneity, also considering the small total number of included trials. However, the primary aim of the present meta-analysis was to specifically evaluate the global functioning after 12 weeks of combined treatment versus antidepressant alone throughout different diagnoses. Moreover, some sensitivity analyses were performed to separately consider MDD patients.
Similarly, in the included studies different kinds of CBT treatments have been offered, especially considering specific interventions for major depressive and anxiety disorders. Moreover, fewer sessions were attended in the study by Clarke et al. (Reference Clarke, Debar and Lynch17) in comparison with others (Table 1).
Another limitation is represented by the difference in age ranges among the studies: in fact the article on anxiety (Reference Walkup, Albano and Piacentini18) was different from the others (Table 1) since presented results of children between the ages of 7 and 17 years.
Furthermore, the reported results pertain to the acute treatment outcome (12 weeks) and could not be extended to longer follow-up. In fact at the end of the 36-week treatment, the rate of clinical response in TADS trial was similar across fluoxetine, CBT, and combined treatment, thus indicating that the benefit of the combination of antidepressant and CBT may be limited to the acceleration of the improvement process (Reference March, Silva and Petrycki26). Furthermore, baseline scores for each scale have not been included in the analyses.
Moreover, even if fluoxetine, the only SSRI currently approved by the US Food and Drug Administration for the treatment of adolescent depression, is the molecule administered in three of the five studies, some considerations should be taken into account in the evaluation of results: (a) the use of a different molecule (sertraline) in the study on anxiety disorders (Reference Walkup, Albano and Piacentini18), (b) the inclusion of venlafaxine in the TORDIA trial (Reference Brent, Emslie and Clarke15), and (c) and the fact that in the study by Clarke et al. (Reference Clarke, Debar and Lynch17) the administered molecule was not specified but only the antidepressant class (SSRIs).
Finally, in the whole studies, the combined treatment with CBT has been evaluated and the results cannot be generalised to alternative psychosocial treatments.
In conclusion, combined treatment seems to be more effective than antidepressant treatment alone on global functioning and general improvement in adolescents with major depressive and anxiety disorders. Future studies should focus on different psychotherapeutic treatments, such as interpersonal psychotherapy.
Acknowledgements
This study was performed in the context of the multi-centric project ‘Prescription, Safety, Pharmacogenetics and outcome of Psychotropic Drugs in Childhood and Adolescence: a Cohort Study' (PREMIA). The PREMIA group includes the following regional coordinators and consultants: G. Colasurdo, M. Fabiani, L. Ferraresi, E. Franzoni, A. Frigerio, F. Masina, S. Palazzi, S. Parisi, A. Parmeggiani, A. Preti, C. Ricciutello, M.B.L. Rocchi, G. Sartorio, D. Sisti, A. Squarcia, S. Trebbi, D. Turchetti, P. Visconti, and A. Tullini.
