When you read an article in a medical journal, all you have to go on is what you’re told in the article itself. If you watched Dr. Healy’s commentary [background music…], you know that this 11 page article represents 77,000 pages of data locked away in some data archive out of sight,
a compression ratio of 7,000:1! And if you question an article, there’s no real way to answer your questions without that data. In this case, because of a legal challenge in 2004, the Clinical Study Report
has been available on the Internet for a long time. It’s a 528 page document used to submit the paper to regulatory agencies. Over the years, many have read it over and over and found further things to fuel our contention that the original article reached an indefensible conclusion. But all that really did was further refine suspicions. It didn’t prove a thing:
In 2012, GSK finally posted the actual data
[Appendices B, C, and D] as they had agreed to do in 2004, and so the numbers were there to see. So many numbers!
And the only way to analyze them would be to hand copy them into some electronic format that could be input into a statistical program for reanalysis. I had a shot at that [cataloged in the lesson of Study 329: an unfinished symphony…
], but there were so many numbers!
Too many. I did enough to gain the conviction that this study was as far off the mark as it appeared. But it was only when we got the raw data in an electronic format that we could really do a complete analysis like the one we are publishing. I hasten to add that the form that data came in was a real challenge – a restrictive remote desktop that made the data manipulation very difficult.
The safety analysis required more data access. The transcribed numbers in the IPD tables for the rating scales were fine for the efficacy part, but the IPD version of the Adverse Events weren’t enough. We needed to look at the actual forms filled out during the study by the blinded clinicians and raters to approach the level of nuance needed to reach any conclusions about harms.
Our article isn’t really about Paxil Study 329. People like reporters Shelley Jofre of BBC’s Panorama, Alison Bass who wrote Side Effects, or legal actions from patients and governments brought it to the fore. The courts have levied punishments and record breaking fines already. And our group has been able to add a counter to the original article in the JAACAP which still sits in our libraries un-retracted.
The broader point of our article is that physicians and the patients we advise have an absolute right to look at the raw data behind the abbreviated proxies that appear in our literature as journal articles. When we have that kind of access, the playing field is level and the profession has the necessary means to join in the kind of checks and balances system that keeps people honest. Our paper is an example of how we think that information should be presented. Further, the medical profession has an absolute obligation to do whatever it needs to do to insure that the information we pass on our patients as scientific transcends other influences – including commercial profit or the academic advancement of the authors.
It’s a paradox that many of the authors who have lent their reputations and the reputations of their universities to these jury-rigged Clinical Trials preach a gospel of evidence-based medicine. And these questionable Clinical Trial articles are certainly filled with icons representing the tools of science – graphs, tables, p-values, standard deviations, etc. But they hide the only basic scientific tool we will ever have – the carefully gathered primary observations we call data. The real evidence never makes it into the courtroom…