This article has Open Peer Review reports available.
Even more suicide attempts in clinical trials with paroxetine randomised against placebo
© Aursnes et al; licensee BioMed Central Ltd. 2006
Received: 15 August 2006
Accepted: 28 November 2006
Published: 28 November 2006
Following our previous publication we have received critical comments to our conclusions as well as new data that are strengthening our findings.
With the new data, 11 suicide attempts among patients on paroxetine against 1 among patients on placebo, we found with a Bayesian technique that the posterior probability that medication with paroxetine is associated with an increased intensity per year of a suicide attempt is from 0.98 to 0.99, depending on the prior.
We found that the comment to our article by GSK representatives contained errors, misunderstanding and unwillingness to accept Bayesian principles in the analysis of clinical trials.
We were in our previous publication, with preliminary data and a Bayesian approach, able to raise a concern that suicide attempts might be connected with the use of paroxetine. This suspicion has now been confirmed.
Last year we wrote a paper "Suicide attempts in clinical trials with paroxetine randomised against placebo"  that hit the front pages of newspapers worldwide [2, 3]. Our publication demonstrated an increased intensity of suicide attempts per year when using paroxetine compared to placebo, and caused GlaxoSmithKline (GSK) to come up with a comment . Since then GSK has provided additional data to the American Food and Drug Administration (FDA), as the agency required new documentation on paroxetine. This also resulted in a Briefing Document from GSK  in which they admit that there is an increased risk for suicide attempts associated with paroxetine.
In addition to calling into question the statistical method applied, the authors of the GSK comment denied our findings by disputing the selection of data. They state that "analysis was based solely on early data submitted to regulatory authorities in 1989. The studies analyzed included only 916 paroxetine-treated patients. Today we have data from more than 13,000 adult patients treated with paroxetine in well-controlled clinical trials." Interestingly, GSK has now in the Briefing Document reduced the corresponding number to 3455 and also admitted that they previously had several patients with a history of suicide attempts from an earlier trial placed in the treatment and placebo groups. With such a case history these patients were more likely to attempt suicide again. They should obviously not have been included since they blurred the results. This can be described as 'data-drowning', and is at least poor methodology.
One item in the comment by GSK concerns two cases that they believe we had misplaced. One patient (no 1.12.037) was claimed to show only suicidal ideation but has by GSK in the Briefing Document been re-classified to represent a suicide attempt. The other patient (no 04.02.056) was from a cross over continuation study. We do not agree that this patient should have been excluded.
GlaxoSmithKline state in their comment to our article that the risk may be closely related to the point of time observed in a study, and further that they would expect to have 6 attempts in the treatment group out of a total events of 8. The number is more precisely 5.79. Hence, our number 7 represents a 20.9 % increase compared to what is expected. They apply a Fishers Exact Test, which in our opinion is a very coarse method. Both in their earlier report in which they applied the Mantel-Haenszel test, and in their new analysis in the Briefing Document  where they applied a so-called exact method, they do not take the exposure time of the patients properly into account, as we have done.
In the comment to our article it was claimed that we should have compared a hypothesis of an increased number of suicide attempts to one in which the rate is identical for the treatment group and placebo group. This hypothesis is meaningless. No one believes that the rates for these two groups are exactly equal. Secondly, we do not in any way attempt to do any form of hypothesis testing. We are merely computing the probability that medication with paroxetine is associated with an increased intensity per year of suicide attempts. GSK questions our statement that '...the data strongly suggest that the use of SSRIs is connected with an increased intensity of suicide attempts per year'. We feel that this is all about understanding and accepting the Bayesian approach as a way of analysing data. Recently FDA issued a drafted guidance for using Bayesian statistics in clinical trials . They write 'Bayesian statistics is a statistical theory and approach to data analysis that provides a coherent method for learning from evidence as it accumulates.' Hopefully FDA's guidelines will contribute to an increasing understanding of the Bayesian approach.
We conclude that irrespective of the various prior assumptions, we now see clearer that the data strongly suggest that the use of antidepressant drugs is connected with an increased intensity of suicide attempts per year, not only among young people, but also among adults. Moreover, information from studies on antidepressants and other drugs must be made open for researchers and presented in the most informative way possible. We believe that this would create an open minded, trusting dialogue between consumers, manufacturers and scientists working towards the same goal: finding the best and safest treatment for the patients.
- Aursnes I, Tvete IF, Gaasemyr J, Natvig B: Suicide attempts in clinical trials with paroxetine randomised against placebo. BMC Medicine. 2005, 3: 14-10.1186/1741-7015-3-14. doi:10.1186/1741-7015-3-14View ArticlePubMedPubMed CentralGoogle Scholar
- Hawkes N: Top selling drug linked to increased suicide risk. [http://www.timesonline.co.uk/article/0,,2-1741916,00.html]
- Boseley S: Drug 'can trigger suicide in adults.'. [http://www.guardian.co.uk/uk_news/story/0,3604,1553676,00.html]
- Modell J: Paroxetine association with suicide attempts overstated. [http://www.biomedcentral.com/1741-7015/3/14/comments]
- Briefing Document. 2006 April 5th., [http://www.gsk.com/media/paroxetine/briefing_doc.pdf]
- Guidance for the use of Bayesian statistics in medical device clinical trials – Draft guidance for industry and FDA staff. [http://www.fda.gov/cdrh/osb/guidance/1601.html]
- Aursnes I, Tvete IF, Gaasemyr J, Natvig B: Clinical efficacies of antihypertensive drugs. Scand Cardiovasc J. 2003, 37: 72-79. 10.1080/14017430310001186.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-244X/6/55/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.