The article must look and feel like a scientific study using appropriate logic and format.
The article must be based on the genuine data set of the study.
The conclusion of the article must confirm the original hypothesis or some acceptable alternative…
- The data set does not actually support a desirable conclusion [or, perhaps, any conclusion], and specifically doesn’t actually support the conclusion of the article.
The kind of ghost-writer we’re talking about is tasked to produce a particular type of illusion. Here are the rules:
Whatever device is used to make the article confirm the original hypothesis or some acceptable alternative, it must have a ready explanation that it was not what it actually was – a trick. This is called PLAUSIBLE DENIABILITY.
I intentionally left out one of the rules for a ghost-writer of the type being discussed here:
Most of us learned about PLAUSIBLE DENIABILITY from the political arena. Karl Rove says, "I did not leak her name" referring to Valerie Plame, the CIA Agent he outed to a reporter to discredit her husband Joseph Wilson. What he actually said to the reporter when he outed Ms. Plame was "Wilson’s wife" [not her actual name]. President Clinton cracked us all up when he stretched the limits of plausibility, "I did not have sex with that woman" [oral sex apparently didn’t count as sex in Hope Arkansas]. So, for the ghost-writer, the playing around with numbers needs to be explainable, even if the true authorship is revealed.
The paroxetine 352 bipolar trial: A study in medical ghostwriting
by Jay D. Amsterdam and Leemon B. McHenry
International Journal of Risk & Safety in Medicine 2012 24:221–231.
"… Many of the named authors on the published article had little or no direct involvement in the design, daily conduct, data analysis, or writing of the initial manuscript drafts. In fact, some of the authors were only selected for this role once the ghostwriters began to draft the manuscript from the final study report or a summary provided by GSK. It appears from the available evidence that GSK and STI had originally chosen Dr. Laszlo Gyulai, then Assistant Professor at the University of Pennsylvania, as the paper’s first author. However, Dr. Gyulai was subsequently removed from this position by GSK and replaced by two other authors who were assigned by GSK to the first and second positions on the paper. The evidence also indicates that the final GSK-assigned authors on the published article never reviewed or even saw preliminary drafts of the paper, and only saw the final edited manuscript just prior to final acceptance by the American Journal of Psychiatry."
Nemeroff, the paper’s first author, says that the data used withstood rigorous peer review in a process that sent the paper back to the authors for revisions several times. "Right in the abstract under ‘results’ we report that ‘Differences in overall efficacy among the three groups were not statistically significant’," he says. "I don’t know how much more straightforward we can be than that. "He adds that "with a 2011 magnifying glass, obviously one would have included in the published paper the use of an editorial assistant". Still, he says: "All [STI] did was help collate all the different authors’ comments and help with references. We wrote the paper."
Now, we’re in a position to guess why this article is missing the expected graph of the treatments and placebo over the timeline of the study. There was no separation and that’s clearly stated throughout the article. So why leave it out? In the Protocol, they even say they’re going to show it [last sentence]:
My guess is that if they showed us their advertised graph, they’d be expected to show a similar graph of the High Lithium Strata versus the Low Lithium Strata, the artificial post hoc comparison that they are calling significant, and we’d be able to immediately tell that it’s not a meaningful conclusion. That’s how this kind of ghost-writing works. In Paxil Study 329, they didn’t show us a graph that makes it immediately apparent that a value they base their significance on is an outlier, a spurious value that didn’t characterize the data accurately [our real scientists…]. GSK can easily disprove my speculation by releasing the raw data from Paxil Study 352 like they were forced to do with Paxil Study 329. If I’m wrong, I’ll publish my apology in all capital letters, in bold type.
Just one more comment about this kind of ghost-writing. It’s hard work, fulfilling the requirements of all those rules. One has to run the numbers all kinds of different ways until you find something you can use. Then you have to figure out how to present it as credible. Then you have to jury rig what’s omitted and included all the while insuring PLAUSIBLE DENIABILITY. It’s hard not to leave behind signs of all that activity. This article has something that might be such a sign. In the quote from the Protocol above, it says, "… no data are carried forward to estimate missing data points." And yet in the Results Database on the site, the headings for the Data Tables say:
LOCF stands for Last Observation Carried Forward, which has nothing to do with the table as it’s presented. I’d speculate that’s left over from some former use of the table as they searched for something of significance. Who knows? It’s sure an odd heading for the table as it is now, particularly with them saying they weren’t going to do that in the Protocol.