In the present effectiveness study, we investigated the course of parent- and adolescent-rated behavioral and emotional symptoms of mental disorders in a sample of clinically referred adolescents with depressive disorders undergoing routine outpatient CBT. The treatment was delivered by psychologists and educationalists with advanced training in CBT in an outpatient clinic. Changes were analyzed for different subsamples and for the total sample, and were compared to a historical control group of patients with depression who had received treatment as usual. Furthermore, the clinical relevance of these changes was assessed and improvements during treatment were compared between patients who received CBT and patients who received CBT plus antidepressant psychopharmacotherapy.
The results revealed statistically highly significant reductions of depressive symptoms and symptoms of other mental disorders in the total sample, as rated by patients and their parents. Mostly large symptom reductions emerged in the subsample which displayed elevated symptom scores on the respective analyzed scales at the start of treatment. Additionally, correlations between parent- and adolescent-rated symptom reductions as well as between parent and adolescent ratings for a broad range of behavioral and emotional problems were moderate and statistically highly significant. This suggests that in terms of their ratings concerning symptom reductions during treatment, adolescents and their parents show moderate levels of agreement with one another.
Our analyses on clinical significance show that a larger share of the sample was clinically normalized at the end of treatment, but more than half of the sample remained in a clinical range. In sum, our hypotheses 1 and 2 can be mostly confirmed, as many youngsters showed a statistically and clinically significant reduction of mental disorder symptoms during treatment. Future analyses of differential effects should be conducted to determine which patients benefit from the delivered treatment, and how the treatment may be improved in order to reduce the proportion of adolescents who are rated in a clinical range by either informant (parents, self) at the end of treatment.
As the present effectiveness study did not use a randomized controlled design, a historical control group was used to control for regression to the mean and unspecific treatment effects. The study by Weisz and coworkers [25], which investigated therapies of youth with depression, is of interest in this regard, as patients and therapists were recruited within routine care settings.. Nevertheless, it is important to take into account some relevant differences between our sample and the historical control sample when interpreting the present findings: The historical control group was (1) on average about four years younger, (2) less clinically impaired at the start of treatment (nearly half of the sample with MinDD or DD compared to 12.2% of our sample), (3) differed in terms of in- and exclusion criteria and had a much shorter treatment than the routine CBT treatment investigated in our study (average: 20 treatment sessions in 40 weeks vs. 44 treatment sessions in 73 weeks) and (4) a substantially smaller sample size. Due to these differences, we did not conduct direct comparisons in terms of testing for statistically significant differences between the two samples. When comparing this historical control group which received treatment as usual with a subsample of our participants with similar T-scores on the CBCL scales at intake (patients rated in the clinical range at start of treatment), medium to large net effect sizes in favor of our routine CBT group emerged in self- and parent rating, thus confirming our third hypothesis. One possible explanation for the superiority of our routine CBT treatment is that compared to the brief CBT training in the study by Weisz and coworkers (2009), the therapists who provided our routine therapies had much more CBT knowledge and experience, with two to five years of CBT training. On the other hand, it has to be kept in mind that the treatment length and intensity in our study was much higher in terms of substantially more treatment sessions. Therefore, it cannot be ruled out that the higher effects may be attributable to the different treatment intensity. Future studies on differential effects will have to investigate the potential influence of variables such as treatment intensity/ duration or the level of CBT training on symptom reductions. Despite the limitations in terms of the comparability of these two groups, through the use of a historical control group to monitor regression and unspecific effects, we can assume that the symptom reductions found in the present study are not solely attributable to developmental trends or regression effects.
It is difficult to draw comparisons between the results of our observational study and previous published studies which included routine therapy, as study designs, treatments, therapists and sample characteristics differed. Nevertheless, it seems important to review their findings. Three reviews/ meta-analyses are of special interest in this regard. First, the meta-analysis by Michael and Crowley [17] examined 15 controlled studies and reported a small pre-post effect size of d = .37 in self-rating for any form of control condition (mainly waiting-list or no-treatment). Second, the network meta-analysis by Zhou and coworkers [10] analyzed 52 studies and found no treatment effect at all in self-rating − all forms of routine treatment were as effective as a waiting-list or no-treatment condition. The pre-post effect sizes found in our total sample were large (d = .82 in self-rating and d = .81 in parent rating; and in subsamples with elevated symptom scores at the start of treatment on the respective analyzed scale: d = 1.20 in self-rating and d = 1.05 in parent rating), indicating a statistically significant, medium to large reduction of depressive symptoms during routine CBT. When relating these findings to one another, it can be concluded that routine CBT is more effective than waiting-list, no-treatment or other forms of routine psychotherapy if used as control conditions in RCT studies. However, in a recent review and meta-analysis, Bear and coworkers [20] investigated different forms of routine therapy that were delivered in unspecified specialized outpatient treatment units with more than half of the studies using a non-controlled pre-post design. Overall, the authors found a large reduction of depressive symptoms across all forms of routine therapy and across different raters (d = 0.89). This result is comparable to the pre-post effect sizes found in our total sample. Moreover, in self-rating, a recovery rate of 40% was found, which, depending on the respective instrument, is nearly equivalent to the recovery rates in our sample (YSR Total problems: 45.0%; SBB-DES total score: 38.3%). Therefore, our results are in line with the limited previous research on all forms of routine therapy, and add important knowledge to the research field by identifying large symptom reductions during CBT interventions delivered in a routine care setting. In this respect, it is important to note that especially in terms of depression, unspecific treatment effects based on patients’ expectations are particularly high. As such, specific methods potentially explain only a smaller proportion of the total symptom reduction (see for instance [42]. When relating our results to the interesting work of Bear and coworkers [20], it has to be kept in mind that the studies included in their meta-analysis were mostly of poor methodological quality, for instance key information was missing, and some studies did not assess clinical diagnoses using structured clinical interviews or had limited sample sizes. Moreover, the outcome scores of all included studies were aggregated into one single score. Finally, when taking into account our subsamples of patients with elevated symptom scores on the respective scale at the start of treatment, larger effect sizes were found, which exceed those found in the aforementioned meta-analysis. This might indicate a potential superiority of routine CBT compared to other forms of routine therapy. Future studies using an RCT design and including an active control condition should focus on variables that might influence treatment effectiveness, such as treatment components, dosage, rater effects or sample characteristics.
Compared to earlier, highly controlled efficacy studies reporting pre-post effect sizes of d = 1.23 [17] and between-group effect sizes (mostly based on self-ratings) of d = .34 to d = 1.27 [14, 15], our effect sizes within the total sample are clearly smaller. However, the overall effect size found in the recent meta-analysis by Weisz and coworkers [9] was clearly smaller (d = .29). Nevertheless, the difference between these highly controlled efficacy studies and our results is presumably due to the fact that the efficacy studies differed considerably from our effectiveness study in terms of patients and treatment characteristics. The efficacy studies mostly recruited samples specifically for the purpose of the respective studies, using very strict inclusion and exclusion criteria. Moreover, the therapists were intensively trained for the studies and received a large amount of supervision. By contrast, our sample was very heterogeneous with regard to symptoms and comorbid disorders, and comprised clinically referred young people with serious clinical impairments. As it is well known that comorbidity may negatively affect treatment outcome (e.g., [23]), this heterogeneity constitutes a major challenge within effectiveness studies. Although every patient in our study had a depressive disorder, depending on the rating scale, only 44.7 to 82.8% of the total sample lay in the clinical range at the start of treatment on scales assessing depressive symptoms in parent rating, and between 26.3 and 76.4% in adolescent rating, which might be attributed to factors such as dissimulation. For this reason, the scope for symptom reduction during treatment is reduced on these scales. In this context, our analysis of the subsample with elevated symptom scores on the analyzed scale at the start of treatment is of special interest: Large effect sizes were found, which are comparable to the pre-post effect sizes reported for the RCTs assessing manualized CBT that mostly relied on self-rating [17]. Our study therefore adds important knowledge to the research field by demonstrating patient-rated symptom reductions under CBT delivered under routine care conditions that are comparable to highly controlled studies. Due to considerable differences between raters, researchers are increasingly calling for the inclusion of multiple informants in order to optimize assessment objectivity (e.g., [9, 22]). Nevertheless, most of the published studies relied on self-ratings. One particular strength of our study is that besides the assessment of self-ratings, we also investigated parent-rated symptom changes, and our results show that the reported reductions of depressive symptoms were high but slightly inferior in parent rating compared to self-rating (d = 1.05 vs. d = 1.20). Possibly, these findings may be attributed to the fact that several depressive symptoms relating to depressive thoughts and feelings can be described more accurately by the patients themselves than by their parents.
Our comparisons of patients who received CBT alone with patients who additionally received antidepressant medication revealed no group differences for most variables. However, adolescents with additional antidepressant medication reported more internalizing problems and more depressive symptoms at the end of treatment (small effect). This suggests that the main effects found in our total sample might not be attributed to the effects of the pharmacological interventions. However, we cannot confirm our hypothesis that stronger treatment effects can be found in patients with additional antidepressant medication. Although for the most part, there were no differences between these two groups at the start of treatment, one possible interpretation is that patients with an indication for additional pharmacotherapy may need this combined treatment in order to attain treatment effects that are comparable to the effects of CBT in patients without an indication for additional pharmacotherapy.
Our study comprised an average treatment duration of 17 months, and an intensity of almost 43 sessions, thus clearly exceeding previous RCTs and naturalistic studies. Therefore, future studies should be conducted to clarify whether less extensive routine CBT interventions are similarly effective.
The lack of a control condition for the total sample constitutes the most important limitation of our observational study. As such, we are unable to rule out whether the observed changes may be caused by confounding factors other than the treatment, for instance natural developmental trends. However, several studies have demonstrated the stability of mental disorders in adolescents over one to three years. For example, a representative cross-sectional study comprising nearly 3000 4-18-year-olds in Germany found no significant decreases in behavioral and emotional problems (assessed using the CBCL and YSR) with increasing age over a period of two to three years [43].
Although the therapists (in training) in the present study were guided by supervisors in terms of implementing the CBT, and had regular discussions about the treatment sessions, we did not formally assess treatment integrity. Furthermore, while the therapies were performed in a routine care setting in terms of a university outpatient clinic, and were delivered by therapists with advanced CBT training, future studies should examine whether this type of therapy differs from therapy delivered by therapists in outpatient units or private practice under routine care conditions. A further limitation pertains to the representativeness of the analyzed data: Only patients with at least 10 treatment sessions were included in the analysis, and due to missing data, it was not possible to include every treatment in the analysis. When comparing patients included in the analysis with those who had incomplete data, we found that the included patients were significantly less impaired at the start of treatment, and that therapists rated the treatments as more effective, although the differences between groups were small. In terms of the comparison between patients with fewer than 10 appointments (brief counseling) and those with longer treatments, we found that the two groups were comparable with respect to sociodemographic factors and clinical impairment. However, therapists rated the group of excluded patients with brief counseling to be less cooperative, to show a minor treatment success for the overall situation, and to have less improvement in global functioning in terms of treatment dropouts. Therefore, it cannot be ruled out that our results overestimate the effectiveness of routine CBT in youth with depressive disorders, even though after imputing missing data, if at all only very small reductions of effect sizes were found. Future studies should thus examine the most common reasons for treatment dropout in order to analyze how treatment dropouts could be reduced. Finally, additional specific instruments and additional raters, such as teachers, should be included. To assess the stability of the changes observed during treatment, follow-up assessments are needed.