An important controversy has been raised by a recent high-profile meta-analysis, which has focused on the effectiveness of antidepressants in children and adolescents with depression. This meta-analysis is now critically discussed in a recent overview. The two articles – and the accompanying editorials – come to very different conclusions, which leave the clinician at first helpless.
The meta-analysis was already published in August 2016 by a Chinese consortium in cooperation with major international methodologists around Andrea Cipriani of the University of Oxford in the Lancet (Cipriani et al., Lancet 2016; 388: 381-390). The authors investigated all (published and unpublished) randomized, double-blind studies comparing antidepressants to placebo or an active comparator for the acute treatment of depressive disorders in children and adolescents. There were, however, only very few direct comparative studies of two antidepressants. Primary endpoints were efficacy and tolerability, a so-called network meta-analysis was carried out.
The analysis was based on 34 studies with a total of 5,260 patients comparing 14 different compounds: amitriptyline, citalopram, clomipramine, desipramine, duloxetine, escitalopram, fluoxetine, imipramine, mirtazapine, nefazodone, nortriptyline, paroxetine, sertraline and venlafaxine. With regard to efficacy, the authors found that only a single substance, fluoxetine, was significantly superior to placebo with a mean effect size. Fluoxetine was significantly better tolerated than duloxetine and imipramine. Patients treated with imipramine, venlafaxine, or duloxetine dropped-out due to adverse events significantly more often than those who received placebo. Venlafaxine significantly increased the risk of suicidal ideation and behavior. However, the authors limit the generalizability of their results due to the considerable methodological deficiencies of the studies underlying the analysis. They conclude that if the benefit / risk profile of antidepressants is taken as a basis for depression treatment in childhood and adolescence, these substances would not provide a clear advantage for children and adolescents. If drug treatment is indicated, fluoxetine is probably the best option.
Jon Jureidini from the University of Adelaide, Australia, concludes in the accompanying editorial (Jureidini, Lancet 2016, 318: 844-845) that there are different reasons for the supposition that the data underlying the meta-analysis is probably still too positive. Effectiveness and tolerability would probably be overestimated, and the risk of suicide would be underestimated, so that there would be little evidence to justify the administration of antidepressants to adolescents – and even less for administration to children. It would be particularly important to criticize the fact that the authors did not have access to the individual data of the patients in order to create a more differentiated analysis. Jureidini states that this would be essential for future clinical trials, as only then an objective and high-quality assessment of the efficacy and safety of a therapeutic intervention is possible.
In an article now published in the American Journal of Psychiatry, John Walkup of the Department of Psychiatry at Cornell University, New York, picks up the hypothesis (Walkup, Am J Psychiatry, 2017, 174: 430-437) that most of the industry-sponsored studies on which the meta-analysis of Cipriani et al. is based, are not “negative” studies (i.e., studies that show no superiority of the antidepressant over placebo), but “failed studies”, which means that they are not suitable for the purpose of determining the effectiveness (and tolerability) of the test compounds at all. However, he restricts his analysis to newer antidepressants (thus excludes tricyclic substances). He claims that there are only two studies sponsored by the National Institute of Mental Health (NIMH), which are methodologically so accurate that they could be used for a meta-analysis. Both these studies tested fluoxetine against placebo and both showed superiority of the antidepressant. The meta-analysis of Cipriani would come to false conclusions, because it is based on countless methodologically deficient and therefore worthless studies. For example, the rate of placebo responders is significantly higher in industry-sponsored studies (50-60%) than in the two NIMH-sponsored studies (35%). The efficacy of antidepressants in the case of obsessive-compulsive disorder and anxiety disorders would suggest that these substances also act in depressive disorders.
Critical questions must be asked. Is the inference of obsessive-compulsive and anxiety disorders really justified when at least the newer antidepressants have only two studies that meet methodological standards? Is it possible to conclude from the fact that numerous studies were “failed studies” (if one follows John Walkup in this assessment), that antidepressants are therefore effective in depression? Probably not. There is just no evidence.
There is also the question of the value of the numerous efficacy studies for depressive disorders in adults. These studies, most of which are sponsored by the industry, are the basis for approval of antidepressants for the treatment of depression in adulthood, and they have been accepted by the regulatory authorities for this purpose. Is a clinical study, therefore, a methodologically acceptable study, if it shows the superiority of the investigational drug over placebo, but a “failed study” if this proof fails?
As always PDFs of the quoted articles can be requested with me.
This post is also available in: German