Commentary on Z hou et al . (2015): Treating psychiatric comorbidity in adolescents—an important problem

Research on treatments for substance use disorders among adolescents and young adults has, to date, focused mainly on behavioral interventions 1, 2. There have been fewer studies in these younger patients of medications for treatment of substance use disorder, or of treatments targeting co-occurring psychiatric disorders. In this context the meta-analysis by Zhou et al., in this issue 3, is a welcome contribution, providing a useful synthesis of placebo-controlled trials of antidepressant medications for treatments of adolescents/young adults with both substance use disorder and major depression. As the paper notes, co-occurring psychiatric disorders are prevalent among adolescents who seek treatment for substance use disorders, and are associated with greater severity and worse prognosis. This suggests the hypothesis that adequate treatment of such co-occurring psychiatric disorders might also help to improve the outcome of the substance use disorders. The number of studies that could be included in the meta-analysis, and sample sizes of those studies, are relatively small. However, placebo-controlled studies of this type are arduous to carry off for a variety of reasons, including the need to find patients with both disorders, which limits recruitment, and the difficulties of engaging adolescents in treatment. Thus, this sample of studies is valuable. Further, studies of this type are investigator-initiated and publicly funded, so that publication bias, or the ‘file drawer problem’, seems unlikely, as investigators will be motivated to publish their work, regardless of the findings. The authors, with appropriate caution, conclude that there is a modest beneficial effect of medication on depression in these young dually diagnosed patients. Caution is warranted because of the limited sample size and mixed findings. The binary depression response variable shows a significant overall benefit of medication. The raw rates of response (defined in a standard manner as a 50% reduction in depression severity scores) across the studies are 67% (97 of 144) on active medication, compared to 56% (82 of 146) on placebo. The meta-analysis of the depression scale scores, analyzed as continuous variables, is not significant, although the two largest studies in the sample 4, 5 yielded favorable effect sizes (standardized mean difference) in the 0.50 range. Analogous studies among adult patients presenting for treatment of substance use disorders and co-occurring depression have similarly yielded modest effect sizes 6-9. Given these findings, where should we stand in recommending the use of antidepressant medications for young patients with substance use disorders and co-occurring depression? Three salient features of the findings suggest the need for a nuanced approach—the modest size of the antidepressant effect, high placebo response and unclear impact on substance use outcome. A high placebo mood response could indicate that many of the patients' depressions responded to the concurrent counseling-based treatment they were receiving for their substance use disorders. Four of the five studies offered cognitive–behavioral therapy. Placebo response could also reflect diagnostic confusion—what presents as a major depressive syndrome may actually represent either a transient mood syndrome related the turbulence of adolescence or substance-induced mood symptoms that improve as a patient enters treatment and reduces substance use. These considerations suggest the wisdom of starting by focusing treatment on the substance use disorder, perhaps including some behavioral therapy for depression. Clinical judgment would be needed to decide which patients have a depression that is severe enough, or convincing enough by history, to warrant immediate intervention with medication. The lack of effect of medication on substance use outcome also suggests caution. The central hypothesis, that treating depression will improve substance use outcome, implies a mediated effect—treatment improves depression which, in turn, improves substance use outcome. This is a low-power situation, making an effect on substance use difficult to detect. Further, studies in adult alcoholics (not selected for depression) have found that serotonin re-uptake inhibitors may worsen drinking outcome among the subtype of alcoholics with early onset 10, 11. Another layer of complexity is that depression is highly comorbid with other mental disorders, including anxiety disorders [e.g. post-traumatic stress disorder (PTSD) and social anxiety disorder], attention deficit hyperactivity disorder and bipolar disorder, which may actually have stronger associations with substance use disorders 12. In this sense, depression may be a signal indicating the presence of one of these other disorders, which might ultimately be the more fruitful targets for intervention. For example, treatments for bipolar disorder 13, 14 or PTSD 15 have evidence of effectiveness among patients with substance use disorders. There are a number of relevant questions to address in future research. How to improve diagnostic evaluation? What features of a patient's history (e.g. severity or chronicity of depression, family history, other co-occurring disorders) or initial clinical course (e.g. response to initiation of counseling and behavioral therapy) might help to identify those depressed, substance-abusing adolescents most likely to benefit from antidepressant medication? How to improve treatment? Would a combination of antidepressant medications with particular behavioral therapies be more effective? Are some classes of antidepressants more effective than others in this population? So far, only selective serotonin re-uptake inhibitors have been studied. It is tempting to declare that more, and larger, randomized controlled clinical trials (RCTs) are needed. However, given the cost and logistical challenges of such trials in adolescents and young adults, this may not always be realistic. Many of the most relevant questions (e.g. identifying subgroups of patients most likely to respond, comparing the effectiveness of several treatments) would require very large samples indeed. So-called large pragmatic randomized trials, or observational studies, based on patient registries with data capture from electronic medical records and ‘learning health systems’ represent a potential alternative strategy 16, 17. Such approaches are, in theory, capable of yielding large samples rapidly and efficiently. Such designs should be considered, alongside traditional RCTs, in order to move this important field forward. None.