I try not to be boring and tedious, but sometimes it just can’t be helped. This post isn’t yet another commentary on Paxil Study 329, but it’s in the mix. It’s a timeline about all the studies SKB and GSK did on Paxil in adolescents. I think it’s of interest, but it may put others to sleep. So there – you’ve been warned.
Efficacy of Paroxetine in the Treatment of Adolescent Major Depression: A Randomized, Controlled Trial
by MARTIN B. KELLER, NEAL D. RYAN, MICHAEL STROBER, RACHEL G. KLEIN, STAN P. KUTCHER, BORIS BIRMAHER, OWEN R. HAGINO, HAROLD KOPLEWICZ, GABRIELLE A. CARLSON, GREGORY N. CLARKE, GRAHAM J. EMSLIE, DAVID FEINBERG, BARBARA GELLER, VIVEK KUSUMAKAR, GEORGE PAPATHEODOROU, WILLIAM H. SACK, MICHAEL SWEENEY, KAREN DINEEN WAGNER, ELIZABETH B. WELLER, NANCY C. WINTERS, ROSEMARY OAKES, AND JAMES P. MCCAFFERTY
Journal of the American Academy of Child and Adolescent Psychiatry, 2001, 40:762–772.…Conclusions: Paroxetine is generally well tolerated and effective for major depression in adolescents.
An International, Multicenter, Placebo-Controlled Trial of Paroxetine in Adolescents with Major Depressive Disorder
by Ray Berard, Regan Fong, David J. Carpenter, Christine Thomason, and Christel Wilkinson
Journal of Child and Adolescent Psychopharmacology. 2006 16[1-2]:59–75.…Conclusions: No statistically significant differences were observed for paroxetine compared with placebo on the two prospectively defined primary efficacy variables. Paroxetine at 20–40 mg/day administered over a period of up to 12 weeks was generally well tolerated.
Paroxetine Treatment in Children and Adolescents With Major Depressive Disorder: A Randomized, Multicenter, Double-Blind, Placebo-Controlled Trial
by GRAHAM J. EMSLIE, KAREN DINEEN WAGNER, STAN KUTCHER, STAN KRULEWICZ, REGAN FONG, DAVID J. CARPENTER, ALAN LIPSCHITZ, ANDREA MACHIN, AND CHRISTEL WILKINSON
Journal of the American Academy of Child and Adolescent Psychiatry. 2006 45:709-719.…Conclusions: Paroxetine was not shown to be more efficacious than placebo for treating pediatric major depressive disorder.
So in the three studies, none of the five primary efficacy variables [my "five easy pieces"] reached significance [last column, P]. Now notice the column that says either Odds Ratio or Effect Size:
|Strength of Effect
|We all learned about p values, probability, that ubiquitous statistical measure that tells us that the observations in a comparison are probably not do to chance. In clinical trials, separation from placebo is the gold standard, but it tells us nothing about the strength of the effect of a drug. There are some other indices that estimate strength of the effect – the odds ratio and the effect size. They are often not in the articles when they’re published which stick with p, but they ought to be there. The odds ratio is for categorical variables like response rate. A value of 1 means there was no strength over placebo. We’d like to see an odds ratio like 3 to say the drug was strong. The effect size is for continuous variables like fall in the scores on a rating index. The usual cut-off is 0.25. Anything smaller than that is trivial.|
|I’ve calculated the strength of effect for the five primary efficacy variables with the 95% Confidence Intervals. These numbers should be present in every Clinical Trial article. They’re easy to calculate and are clinically meaningful. So Take-Home number one from this post is to calculate them yourself if they’re not provided in the article. As you can see, even if Paxil had separated from placebo, the more important fact is that Paxil doesn’t have a strength of effect in adolescents that would suggest clinical usefulness. These strength of effect values are the ones used in the meta-analyses like those done by the Cochrane Collaborations.|
I’ve previously said more than enough about why Paxil Study 329 was reported as positive [again see the lesson of Study 329: an unfinished symphony…], but I’ve never taken the time to look at the timeline as it unfolded. The figure below is the simplest version I could come up with:
Paxil came to market after FDA Approval for Adult Major Depressive Disorder in 1992 [chasing Prozac, Celexa, and Zoloft]. Not long after, SmithKlineBeecham began the process to get approval for its use in children and adolescents with Study 329 which started in April 1994. Because of recruitment problems, they doubled the number of centers and finished the trial in March 1997. 329 was a trial in the US. There was also an International Trial  begun a year later. When the results came in they were disappointing [from their memo]:
But they pressed ahead and published the infamous JAACAP article in July 2001. Study 377 stayed in the file drawer. Notice on the timeline that they commissioned a third trial  while Study 329 was being written up [lead by three of the Study 329 authors]. I see that as "one for the road," hoping they could get a positive to submit with 329 to the FDA. But it was a bust [Placebo beat Paxil]. Shortly before Study 329 was published, GlaxoWelcome acquired SmithKlineBeecham and GlaxoSmithKline [GSK] came into being:
|SmithKlineBeecham became GlaxoSmithKline
|It’s tempting to say that GSK inherited the Paxil problems from SKB [see comment]. All of these trials were started by SKB. SKB decided to publish Study 329 and squash Study 377. But when I look at that timeline, SKB may have started the ball rolling, but it gathered no moss in the hands of GSK. GSK knew that 377 and 701 were busts. They couldn’t have not known that 329 was a jury-rigged mess. In case they didn’t know, the FDA told them when they submitted it for a Pediatric Extension, Jon Juriedini told them in his 2003 letter to the JAACAP, and Eliot Spitzer, the NY Attorney General told them when he successfully sued them and forced them to post study results for all three studies in 2004. And Efficacy wasn’t the only problem, there was a strong signal for suicidality as an Adverse Effect in these studies, something GSK published at the end of the drug’s patent life, not as soon as they knew it [Black Box Warning]. And they still drug their heels on publishing 377 and 701 until their patent was at its end. No wounded victims at GSK – they were right in there squeezing every Paxil sale out of the market right up to the end. That is Take-Home point number two of this post – GSK was guilty and earned their $3 B fine that included Paxil among their transgressions.|
The Lancet piece is about the analyses based on individual patient data [IPD] from similar trials after trial completion. We noted that among other advantages, access to IPD facilitates “more thorough data checking” and identification of “missing information”. This includes “checking on whether trialists are cheating”, and some IPD analyses have detected this. In deciding how best to make IPD available for data and other checks and reliable re-analyses the logical starting place is the substantial experience acquired during nearly three decades of IPD analyses done by various collaborative trialists’ groups. That experience is important not only in illustrating the advantages of IPD analyses, but also in showing how the privacy of individual patients can be assured.
He is suggesting, I think, that the trialists, the people like himself and those at the Cochrane Collaboration, like Ben Goldacre and his AllTrials Collaborators, the epidemiologists like Peter Doshi and Tom Jefferson, the campaigners like David Healy, etc. should perhaps be the ones to have access to the individual patient data because of  their proven expertise and  their track record at protecting subject confidentiality. That should reassure the people who want access to the raw data for anyone that wants it [like me] and we should stand down. It’s a tempting offer. Lord knows, before I started reading their work, I didn’t know what effect size was, how to calculate the odds ratio, what forest or funnel plots were. In fact, the people in this paragraph are my current heros of the realm, in particular Iain Chalmers:
|an n=1 opinion
|Since medical school, I have venerated the medical literature. The articles come from our highest centers of learning and are written by the leaders in the field – department heads, section chiefs, acclaimed researchers, our best and brightest. The journals are edited by luminaries and the submissions peer-reviewed by expert volunteers. I believed in that process and trusted it. Many of the names on the articles at the beginning of this post come from those upper levels of the profession.|
|And yet when I look over the story in this post, I see no shining signs of integrity glistening back. This is just a story about turning a study with a diminutive glimmer into a shining light for no reason except to sell a product – one that was of little to no help, was toxic to many, and fatal to some. All my former confidence in the process was for naught. I never imagined that such a thing could actually happen – but it did. The main author on Study 329 was the Chairman of Psychiatry at Brown University, one of our best. There are plenty of studies from the current era that are flawed in the same way – plenty.|
|So as much as I respect the trialists pioneers in the previous paragraphs, who knows what the future may bring? Who can guarantee their successors will be any less vulnerable than the academicians on those articles above? So I’m sticking to my guns here. Nobody is going to wander up this mountain to get my vote, but I can tell you what it would be if they did. The only way to assure integrity into the future is to put the raw numbers out there for anyone who wants to look at them. As our critics are fond of pointing out, it’s the physicians that write the prescriptions. So it’s the physicians and the scientists and the patients who have the right and need to know about those numbers. Of course we’ll turn to the experts for opinions. But if I’m responsible for giving the medicine to a sick person, I should be able to see the actual unprocessed data myself. Learning to analyze it is my responsibility. Since I can’t count on the literature to produce the "five easy pieces" I need, I should be able to derive them myself. I think the threat that we can and will do that will keep PHARMA in line. That’s my Take-Home number three of this post – an opinion…|