Restoring Study 329: Paroxetine neither effective nor safe for adolescents

A major reanalysis of Study 329, a Glaxo Smith Kline (GSK) randomised controlled trial (RCT) comparing paroxetine, imipramine and placebo published in the British Medical Journal (BMJ), has concluded that paroxetine is neither effective nor safe for adolescents with depression (Le Noury et al., 2015). This conclusion is in direct contrast to the original finding which reported that paroxetine was ‘generally well tolerated and effective’ (Keller et al., 2001). The reanalysis is part of an initiative called restoring invisible and abandoned trials (RIAT) which encourages formal reanalysis of original trial data to correct prior misinterpretation leading to inaccurate information being used to inform treatment decisions.

Study 329 has long sustained major criticisms. In an accompanying BMJ editorial, Doshi (2015) notes that despite a Food and Drug Administration (FDA) formal review in 2002 reporting that on balance the trial should be considered as a failed trial, a marketing campaign by GSK characterised Study 329 as demonstrating that paroxetine has ‘remarkable efficacy and safety’. Over two million prescriptions for paroxetine were written for children and adolescents in the United States alone. As Doshi (2015) goes on to report the first draft of the paper was not written by any of the 22 authors but by a GSK-hired medical writer and the paper’s lead author – Keller – had been investigated for under reporting financial ties to drug companies in 1999.

The Le Noury et al. (2015) reanalysis only compounds these concerns. They report that paroxetine only produced a positive result when four new secondary outcome measures, introduced following the initial data analysis, were used instead of the initial primary outcomes. Analysing the primary outcome measure (POM) defined in the protocol written in 1994 revealed no group differences. Adverse events were under reported, and the data were only presented if suffered by 5% or more of the patients. There was also under reporting of suicidal thinking and events in the paroxetine arm.

Remarkably, there has been little action taken about these major concerns. The Journal of the American Academy of Child and Adolescent Psychiatry (JAACAP) has not retracted or corrected the paper. The Academy itself has remained silent. Brown University, who employed Keller, appears to have ignored the whole issue; only the student newspaper has run stories that probe Study 329 (Doshi, 2015).

On a more positive note, because of Study 329 and others, the International Committee of Medical Journal Editors (ICMJE) introduced a policy in 2005 requiring prospective registration, including POMs, of all clinical trials into a trial registry (De Angelis et al., 2004). Some journals now require trial registration and declarations of conflicts of interest before articles will be published. On the face of it, these measures would help resolve many of the concerns related to trials such as 329.

However, we recently systematically reviewed all clinical trials (as defined by ICMJE) published between 2009 and 2013 that had been conducted after July 2005 in the top five psychiatric journals mandating prospective trial registration in order to be published. We found that 21 (11.6%) of the 181 clinical trials published in these five journals were unregistered, 61 (33.7%) retrospectively registered and 37 (20.4%) had unclear POMs. Only 30 (16.6%) were prospectively registered with clearly defined POMs (Scott et al., 2015). And this result was for the top five psychiatry journals who stated that trials would not be published without proper registration. While things may have improved, we clearly still have a long way to go to restore confidence in clinical trial data.

What are the implications for clinical psychiatric practice? It is probable that other clinical trials have also selectively reported outcomes, minimised adverse effects and made unjustified claims for the efficacy of their treatments. These results are incorporated into meta-analyses and treatment guidelines. It is therefore likely that we currently overestimate the efficacy of our treatments and underestimate their adverse effects. Obviously initiatives like RIAT may help to re-balance some of the evidence. They should at least put pressure on academic and professional institutions to address allegations of wrong doing. Hopefully, they might also begin a dialogue which questions the scientific basis of much psychiatric practice making us more open to alternative ways of conceptualising and managing psychological distress.

In the meantime, it seems reasonable to state that paroxetine is a poor choice for adolescents with major depression. It is now clear that if it is ineffective beyond its placebo response, it is harmful, sometimes catastrophically and it can be extremely difficult to withdraw from. However, it is still being prescribed. Analysis of New Zealand dispensing data over the last several years shows that more than 400 adolescents are still being prescribed paroxetine each year. The message now has to be that initiating further prescriptions of paroxetine to this group should stop.