the lesson of Study 329: an unfinished symphony…

Posted on Sunday 23 December 2012

My series earlier this year on Keller et al’s 2001 article about Paxil Study 329 was primarily to get things clear in my own mind, but I was also trying the gather the information in one place to send with my retraction request to the American Academy of Child and Adolescent Psychiatry:
Obviously, that series didn’t go to the end of the story. Drs. Juriedini and Tonkin wrote their letter contesting the published results for which they received the Nastygram from Editor Myna Dulcan and a Contemptogram from the authors published with their letter in the JAACAP [naked Emperors, fractious Queens…]. But by 2004, the questions about suicidality as a side effect of SSRIs had become a roar, and the FDA finally required a black box warning to be added to the package insert specifically focusing on the danger to youth. At that point in time, GSK had only published Study 329, but not two other Clinical Trials in adolescents [Study 377 and Study 701] which were equally suspicious. Whether motivated to cover their tracks or for the loftier public health motives they claim, they decided to publish a review of Paxil and suicidality in adolescents covering all three studies. They sent a draft to the Study 329 authors for comment:
Dr. Keller reacted immediately, worried about looking bad because his paper had buried suicidality, and the draft of the review was apparently more forthcoming about his actual data. This fragment below is just his first paragraph, after that he goes on to request specific corrections that remove conflicts with things he’d already said [email]:
He wanted GSK to publish separate papers, not even including his paper with the others, and got pretty heated about the whole issue. As you can easily see [email], he’s in full cover-your-ass mode ["leaving us to the wolves"; "we could look foolish, naive, incompetent, or biased"; "not written with complete integrity and accuracy given the data we had"; "disregarding our responsibility to the proper scientific method, to the public, children and families"]:
Martin Keller knew exactly what he had done with this paper on Study 329. I couldn’t have said it better. It was…
  • foolish, naive, incompetent, … biased
  • not written with complete integrity and accuracy given the data [they] had
  • disregarding [of his] responsibility to the proper scientific method, to the public, children and families
GSK didn’t give him the three separate papers he asked for, but their final article did embed the 329 data with the other two studies, glossing over the inaccuracy of Keller et al’s Paxil Study 329 report.
Evaluation of suicidal thoughts and behaviors in children and adolescents taking paroxetine.
by Apter A, Lipschitz A, Fong R, Carpenter DJ, Krulewicz S, Davies JT, Wilkinson C, Perera P, and Metz A.
Journal of Child and Adolescent Psychopharmacology. 2006 16[1-2]:77-90.

OBJECTIVE: The aim of this study was to summarize results of a blinded review of potential suicidal events and analyses comparing incidence rates between paroxetine- and placebo-treated pediatric patients.
METHOD: One thousand one hundred ninety-one (1191) children and adolescents received paroxetine (n = 642) or placebo (n = 549) during placebo-controlled portions of all acute double-blind trials of paroxetine (n = 5). An expert panel blindly reviewed and categorized all identified cases detected by electronic and manual search of adverse events (AEs), serious AEs, and selected cases as suicidal or non-suicidal behavior. Incidence rates were calculated for suicide-related events and for rating scale items assessing suicidality.
RESULTS: Suicide-related events occurred more often in paroxetine (22 of 642, 3.4%) than placebo groups (5 of 549, 0.9%); odds ratio (OR) 3.86 (95% CI 1.45, 10.26; p = 0.003). All suicide-related events occurred in adolescents of at least 12 years, except for 1 of 156 paroxetine-treated children. All suicide attempts occurred in major depressive disorder (MDD); few suicide-related events occurred in patients with a primary anxiety disorder. Suicide item analyses did not reveal significant differences between paroxetine and placebo.
CONCLUSIONS: Adolescents treated with paroxetine showed an increased risk of suicide-related events. Suicidality rating scales did not show this risk difference. The presence of uncontrolled suicide risk factors, the relatively low incidence of these events, and their predominance in adolescents with MDD make it difficult to identify a single cause for suicidality in these pediatric patients.
If you read that last email several times, it is totally focused on his public image and nowhere says a word about publishing a study that hid dangerous side effects from the reader, much less even mentioning that it claimed fictitious efficacy. I expect we all would like to look good in print, but hopefully, we’d maintain at least some focus on accuracy and that sticky "do no harm" thing. Dr. Keller exhibits none of the basic values of a research scientist, rather appearing to be a full-time self-promoter. He even intimidated GSK, hardly a paragon of virtue themselves. I expect he has done the same thing with the JAACAP. Notice also that Drs. Ryan, Emslie, and Wagner are his main confidentes. There’s more detail to this story, well documented in the comments to my last two posts by Annonymous, and I’m indebted for both the references and the commentary.

I should add that the people I contacted at the American Academy of Child and Adolescent Psychiatry [incoming and outgoing presidents, the ethics committee] were cordial and responsive. Besides the blog series above, I sent them copies of Alison Bass’s book Side Effects, still a classic [used copies, sorry Alison]. I expect that Editor Andres Martin was the decider and under a lot of pressure not to retract this article – whether from GSK, from the authors of the paper, from the editor who accepted the article originally, or from other elements within the Academy. But that said, I find it impossible to accept his comment as the truth about his decision, "the Journal editors found no basis for retraction or other editorial action." The evidence is just too strong and too widely available. I don’t think I’m being a sore loser here. The loser in this story is the American Academy of Child and Adolescent Psychiatry, and ultimately psychiatry itself…

hat tip to Annonymous
    December 23, 2012 | 12:58 AM

    Thanks. When there  is considerable political cost to taking action, and little to no political cost to inaction, inaction is fostered. Keller et al appear comfortable with being non-collegial when they feel their interests are threatened. Broader change cannot be built simply through those willing to martyr themselves or with little to lose. Unless there is a pressure experienced counterbalancing business as usual the current state of affairs will persist.

    “Sooner or later, this whole tawdy saga is going to find its way out of the blogs and courtrooms and into the full light of day.”

    I remain uncertain, but if you are right then may  it be sooner rather than later. 

    In any event, thank you very much for sustaining a forum that has at least kept this discussion alive in the interim. Thanks again for the hat tip. 

    December 23, 2012 | 4:02 PM

    I thank you for this forum as well as the continual postings keeping these topics alive and well–

    Bernard Carroll
    December 24, 2012 | 12:44 AM

    A check on Thomson Reuters reveals that the 2001 report by Keller, Ryan et al in JAACAP has been cited 291 times. That qualifies it as a so-called citation classic. In contrast, the 2006 report by Apter et al in Journal of Child and Adolescent Psychopharmacology has been cited only 16 times. Which would we say contains the best information? No contest. Thus do the academic journals become corrupted in a self sustaining manner.

Sorry, the comment form is closed at this time.