Research paperTreating a broader range of depressed adolescents with combined therapy
Introduction
Among young people, major depression accounts for a substantial portion of the burden of disease (Gore et al., 2011, Merikangas et al., 2010). Furthermore, depression has a variety of adverse effects, both immediately and in adulthood. In particular, depression during adolescence is associated with an increased risk for depression and other psychopathology in adulthood (Fergusson and Woodward, 2002, Jones, 2013, Rutter et al., 2006); an increased risk for self-harm and suicidal behaviors (Fergusson and Woodward, 2002, Gould et al., 1998, Hawton et al., 2012, Thapar et al., 2012); an increased risk for reduced social functioning and work difficulties (Costello and Maughan, 2015, Fergusson and Woodward, 2002, Thapar et al., 2012); and an increased risk of future adverse birth outcomes among women (Nkansah-Amankra and Tettey, 2015). It is clear, then, that depression in adolescents needs to be treated swiftly and effectively.
The ‘Treatment for Adolescents with Depression Study’ (TADS) was a corner-stone clinical trial that tested the effectiveness of standard treatment options for major depression in adolescents, including treatment with the selective serotonin reuptake inhibitor (SSRI) fluoxetine, treatment with cognitive-behavioral therapy, and these two treatments combined (March et al., 2004, Thapar et al., 2012). Statistical analysis of the acute phase indicated superiority of fluoxetine used alone, and of the combined treatment over a pill placebo, as well as of combined treatment over either fluoxetine or cognitive-behavioral therapy used alone; cognitive-behavioral therapy, however, failed to outperform the placebo (March et al., 2004).
Treatment effects were estimated using standard statistical methods, including mixed-effects and logistic regression (March et al., 2004). Whereas such analyses are indispensable to establishing a treatment's general level of effectiveness, they only truly examine outcomes for each treatment in terms of their central tendency, for example their mean value across all subjects within each treatment arm. However, looking at central tendency fails to reveal whether any detected treatment effect exists only on average or across the entire distribution of outcome values (Cleveland, 1994, Koenker and Hallock, 2001, Wilcox, 2012).
The quantile regression framework allows for a more complete look at the differences between the outcome distributions of two treatment conditions. With quantile regressions, the quantiles of an outcome's distribution are modeled as a function of the predictor variable(s), in contrast to the traditional regression framework, wherein only the mean of the outcome is modeled (Beyerlein, 2014, Briollais and Durrieu, 2014, Koenker and Hallock, 2001, Petscher and Logan, 2014). This allows for comparing all quantiles of the treatment condition's outcome distribution against the corresponding quantiles of the placebo condition's outcome distribution. Comparing quantiles is the most informative way of comparing two groups because it shows whether the groups differ in a simple or in a complicated way (Cleveland, 1994, Wilcox, 2012). In the simple case, one group has consistently higher values than the other. In a more complicated case, low, middle, and high outcome values differ by different amounts. It is possible that the groups differ in an even more complicated way, for example with one group outperforming the other at lower outcome values, but being outperformed at higher outcome values. Such complicated group differences cannot be uncovered by comparing the groups’ central tendencies.
Assessing a treatment's effect across the entire outcome range would be especially attractive if the outcome is an index of response (e.g., indexing the degree of change from baseline to the trial's end), since such an analysis would show whether a given treatment has a consistent versus differential effect on low, middle, and high response levels. Thus, rather than examining whether the average response level is higher among subjects receiving active treatment relative to controls, one could examine whether the treatment has some level of effectiveness across the entire range of response levels, including special subject subsets like low and high responders (Beyerlein, 2014, Briollais and Durrieu, 2014, Petscher and Logan, 2014). A treatment that is able to exert a consistent effect across response levels is more clinically valuable than a treatment with a more restricted breadth of effectiveness.
Our current aim was to gain a more complete picture of TADS treatment effects, by re-evaluating them across the entire response range. We did so by analyzing patients’ change in depression severity from baseline to the end of the TADS acute phase as an index of response, using quantile regression models (Beyerlein, 2014, Briollais and Durrieu, 2014, Koenker and Hallock, 2001, Petscher and Logan, 2014). This approach allowed us to analyze both central tendency and outcome extremes, thereby providing more insights into the breadth of the treatments’ effectiveness than traditional regression analyses. Since we are not aware of any comparable previous analyses, the study was exploratory. The research question was whether treatment effectiveness varied between low, middle, and high response levels.
Section snippets
Study design
The current study entailed secondary analysis of TADS trial data (ClinicalTrials.gov, NCT00006286). TADS is a randomized controlled trial that was designed to compare the effectiveness of cognitive-behavioral therapy CBT), a specific SSRI called fluoxetine, and the two treatments combined, all versus a pill placebo for the treatment of adolescents with a major depressive disorder. We analyzed data from the acute phase of the trial, which encapsulated the first 12 weeks of treatment. The data
Results
Table 1 provides the 10th to 90th response level quantiles of each treatment arm. Response levels were rather dispersed. In the placebo condition, the response level on the 10th quantile was 0.067 and as high as 0.61 on the 90th quantile. This corresponds to a reduction of the baseline depression score by 6.7% on the 10th quantile and by 60.1% on the 90th quantile. Similar ranges of response levels were found for CBT and FLX. In contrast, COMB had clearly higher response levels, achieving a
Discussion
In the TADS sample, whereas cognitive-behavioral therapy was found to be no more effective than placebo across response levels, our analyses revealed that combining cognitive-behavioral therapy and fluoxetine was not only effective, but had a more comprehensive effectiveness than fluoxetine used alone. This was indicated by higher AUCs for the combined-treatment effects, reflecting simultaneously that these effects were greater in magnitude and spread out across a broader range of response
Conclusion
We found that, in the TADS sample, the combination of cognitive-behavioral therapy and the SSRI fluoxetine was more effective than either treatment used alone, not just in average effectiveness, but in the breadth of patients in whom it was effective.
Declarations of interest
None.
Contributors
Conceptualization of study: SF and MMK. Data acquisition and preparation: SF. Statistical analysis: SF. Interpretation of data: SF and MMK. Drafting the manuscript: SF. Revising the manuscript for important intellectual content: MMK. All authors have approved the final manuscript.
Funding source
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Acknowledgements
Data used in the preparation of this manuscript were obtained and analysed from the controlled access datasets distributed from the National Institute of Mental Health (NIMH)-supported National Database for Clinical Trials (NDCT, https://data-archive.nimh.nih.gov/ndct/). NDCT is a collaborative informatics system created by the NIMH to provide a national resource to support and accelerate discovery related to clinical trial research in mental health. Dataset identifier: Clinical Trials #2145.
References (79)
- et al.
Comparative efficacy and tolerability of antidepressants for major depressive disorder in children and adolescents: a network meta-analysis
Lancet (London, England)
(2016) - et al.
Predictors and moderators of acute outcome in the Treatment for Adolescents with Depression Study (TADS)
J. Am. Acad. Child Adolesc. Psychiatry
(2006) - et al.
Fluoxetine treatment for prevention of relapse of depression in children and adolescents: a double-blind, placebo-controlled study
J. Am. Acad. Child Adolesc. Psychiatry
(2004) - et al.
Global burden of disease in young people aged 10-24 years: a systematic analysis
Lancet (London, England)
(2011) - et al.
Psychopathology associated with suicidal ideation and attempts among children and adolescents
J. Am. Acad. Child Adolesc. Psychiatry
(1998) - et al.
Self-harm and suicide in adolescents
Lancet (London, England)
(2012) - et al.
Treatment of depression in adolescents with cognitive behavior therapy and medications: a commentary on the TADS project
Cogn. Behav. Pract.
(2005) - et al.
Cognitive-behavioral therapy for adolescent depression: a meta-analytic investigation of changes in effect-size estimates
J. Am. Acad. Child Adolesc. Psychiatry
(2007) - et al.
A comparison of cognitive-behavioral therapy, sertraline, and their combination for adolescent depression
J. Am. Acad. Child Adolesc. Psychiatry
(2006) - et al.
Lifetime prevalence of mental disorders in U.S. adolescents: results from the National Comorbidity Survey Replication–Adolescent Supplement (NCS-A)
J. Am. Acad. Child Adolesc. Psychiatry
(2010)