Verified by 
One of the most incendiary indictments of clinical psychopharmacology trials in children and adolescents was published in the September 2015 Issue of the well respected British Medical Journal (1). This study reanalyzed data from a 2001 paper in the Journal of the American Academy of Child and Adolescent reporting a double blind clinical trial of paroxetine and imipramine compared to placebo in depressed adolescents. The published paper reported that paroxetine was effective and safe for treating depression in adolescents. But the drug company, SmithKline Beecham, (now GlaxoSmithKline) told the FDA that the drug had failed to demonstrate that it was effective in the treatment of depressed adolescents. The FDA agreed with GlaxoSmithKline and did not approve paroxetine for this indication.
The discrepancy between the report in the Journal of the study as successful and the report of the drug company to the FDA of the study as unsuccessful, in part, led a group of psychiatric investigators to undertake the intimidating project of analyzing all of the old data following the methodology of the original research protocol. This effort was performed by a group called RIAT - Restoring Invisible and Abandoned Trials. RIAT believed that understanding the discrepancy between the report of the published paper that paroxetine was successful and the drug company’s report to the FDA that it was unsuccessful in the treatment of adolescent depression might increase transparency of clinical trials.
In evaluating previously reported trials RIAT follows the original protocol for the performance of the trial and notes discrepancies between the trial described in the protocol and the actual performance of the trial. RIAT found numerous discrepancies in the report of the trial to the Journal. In this limited space only a small number can be noted. For more detail consult Study 329.org.
275 adolescents with major depression were studied. The subjects received one of three medications: paroxetine, imipramine or placebo throughout the eight week study. The number of patients in each medication group was approximately the same, and the medication to be given to each patient was selected randomly. Both the patients and the clinical researchers were blind to the medications that were given. Patients were evaluated using a number of measures widely employed to assess depression. One of the measures, the Hamilton Depression Rating Scale, was selected to be the “primary efficacy measure” before the study began. The selection of the primary efficacy measure is critical to the success of the clinical trial because it is this measure that must change to consider the trial successful. The investigators specified that Hamilton Depression Rating Scale scores of the patients on paroxetine needed to change from baseline (pre-medication) more than those of the patients on placebo, or that they needed to drop to a score of 8 or less or that they needed to drop by 50% or more. The Journal paper reported that the Hamilton Depression Rating Scale scores for the patients on paroxetine dropped to scores of 8 or less significantly more often than they did for the patients on the imipramine or the placebo.
A careful reanalysis of the data by RIAT failed to confirm that paroxetine lowered Hamilton Depression Rating Scale scores compared to imipramine or placebo. The RIAT group analyzed the primary efficacy data three different ways, but each time the result was the same: there was no difference. Martin Keller, the senior author on the Journal article, wrote what he called a preliminary rebuttal letter to 329.org and other media outlets challenging many of the conclusions of the RIAT 329 reanalysis study, but Keller did not dispute the RIAT reanalysis of the Hamilton Depression Rating Scale finding, the primary efficacy measure of the study (3).
There is considerable dispute about the secondary measures used in the study. The first issue is whether the group who performed the trial saw the data before they decided which measures they were going to analyze statistically as secondary measures. RIAT accused the Journal authors of having decided upon the secondary analyses after the study was completed. This would be a serious breach of ethics. This was vehemently denied by Keller in his rebuttal.
The RIAT reanalysis of side effects led to the conclusion that paroxetine was not safe as the Keller article claimed. There are a variety of methodological issues surrounding the measurement of side effects in this clinical trial. In the limited context of this 8-week study, paroxetine had more side effects than imipramine and placebo. For example, according to RIAT, there were 8 suicidal threats or behaviors with paroxetine, compared to 3 with imipramine and one with placebo. In a separate analysis, RIAT found 70 serious adverse events with paroxetine compared to 50 serious adverse events with imipramine and 26 serious adverse events with placebo.
The challenges presented by the RIAT analysis are disturbing. The authors of the journal article are well known and deeply respected for good reason. A more fulsome response from these authors to the RIAT group should be delivered to resolve the questions their analyses raise.
This study is a milestone in that it has increased the transparency of the performance of clinical trials. The FDA served its function to protect the public from ineffective and potentially dangerous drugs by refusing to approve paroxetine for the treatment of depression in adolescence.
Post publication review of published studies, such as RIAT has performed, would serve to enhance the integrity of future psychopharmacological clinical trials.
An unwelcome consequence of the RIAT critique is to undermine the mental health community’s confidence of the psychopharmacological treatment of depression. In the wider context beyond this 8-week study, to knowledgeable prescribers, imipramine is well known to be a much more dangerous treatment for depression than paroxetine. It has cardiac side effects that can be lethal if an excessive amount is taken, whereas paroxetine is highly unlikely to lead to completed suicide if taken in excessive amounts such as in an intentional overdose. Furthermore, in a wider context than this 8-week study, placebo is not a safe treatment for depression. Untreated depression is strongly associated with suicide and an array of additional unwanted consequences.
References
BMJ 2015:351:h4320 September 17,2015
Keller, MB., Efficacy of paroxetine in the treatment of adolescent major depression: a randomized, controlled trial. J. Am. Acad Child Adolesc. Psychiatry 40:7, July 2001.
http://davidhealy.org/study-329-mk-hk-sk-and-gsk/
Copyright: Stuart L. Kaplan, M.D., 2015
Stuart L. Kaplan, M.D. is author of Your Child Does Not Have Bipolar Disorder: How Bad Science and Good Public Relations Created The Diagnosis, available at Amazon.Com.
