In the late 1980s, Eli Lilly, when faced with an excess of suicidal behaviors in Prozac trials, set up a trial of Prozac in an interesting group of patients.
These patients had what is often called borderline personality disorder or intermittent brief depressive disorder or recurrent brief depressive disorder. The trial terminated early. Placebo was sweepingly statistically superior to Prozac. The study remained unpublished for four years, only appearing (without its data) when Lilly were facing their first homicide trial — the Fentress case (also known as the Wesbecker case).
The trial had been conducted in London by Dr. Stuart Montgomery. In the early 1990s, SmithKline agreed a study in the same group of patients with Dr. Montgomery (protocol 106). This also terminated early, and again the results were never published.
So why did SmithKline set up another trial (protocol 057) in Europe in a similar group of patients a few years later?
Lots of people could have advised companies to do just this as a means of hiding the bodies. The patients entered into the trial are a group who make regular suicide attempts — several a week is possible. If Prozac or Paxil actually helped, this would be a win for the company. But even if there were more suicide attempts on Prozac or Paxil the studies would still be astonishingly useful — as will become clear.
Hey, presto — the risk is gone
In April 2006, GlaxoSmithKline issued a press release buried in which was a concession that Paxil was linked to a statistically significant increase in suicidal acts in patients with major depressive disorder. But this concession was flanked by the data from protocols 057 and 106 (patients with intermittent depressive disorders) and these data magically made GSK’s problems vanish.
Here are the major depressive disorder figures:
|
Paxil |
Placebo |
Relative Risk |
|
|
No. Acts/ No. Patients |
11/2943 |
0/1671 |
6.0+ |
In this GSK document, protocols 106 and 057 have 298 patients between them and 67 suicidal acts — there are 100 times more suicidal acts in these two protocols than in the entire set of depression trials. There are several different versions of the results, but the results GSK presented in 2006, shown below, show essentially the same number of suicidal acts for Paxil and placebo. Adding these acts into the mix produces a dramatic and unexpected result:
|
Paxil |
Placebo |
Relative Risk |
|
| MDD Trials Acts/Patients |
11/2943 |
0/1671 |
6.0+ |
|
057/106 Acts/Patients |
32/147 |
35/151 |
|
| Combined Acts/Patients |
43/3090 |
35/1822 |
0.7 |
Hey, presto — the risk is gone.
To advise something like this, the advisor has to know the drug causes the problem that the proposed study will hide. It takes a little insider information to work this out but not much…
Would a company do something like this on purpose? The answer here probably hinges on issues of plausible deniability. Someone in the company may have appreciated what was going on, but the CEO or board members may not; protocols 057 and 106 may have seemed quite innocent to them.
Everyone in a company can pass the buck on studies like this. From the time of conception to the time of later use of these studies for this purpose, a decade or more passed in which there was likely a complete turnover of personnel within the relevant divisions in GSK. We depend on feedback from someone inside GSK or other companies on the kind of issues involved in studies 106 and 057. A bright idea like this is not something that is likely to be restricted to psychiatry only — it would be good to have instances of similar studies in other areas of medicine.
This stuff matters
The implications are far-reaching. As Google says in connection with proposed changes to its privacy policy: This stuff matters.
At least one person in protocol 106 was seriously maimed by a suicide attempt, on Paxil, as I understand it. There may have been other serious problems in the other two studies.
Marriage of business and science?
Where does this leave informed consent and ethical approval? Is there an ethics committee or institutional review board on earth equipped to gauge whether a study they have been asked to process is what it seems? How do they begin to raise objections if the true purpose of the study is not contained in the protocol?
Protocols 106 and 057 seriously question whether business can be married to science.
The obvious problem is the conflict of interest. As long as drug companies are the ones who get to design studies about the efficacy and safety of their own drugs, the incentive to skew results will be irresistible. This situation is further complicated by the fact that it is legal for drug companies to bribe prominent physicians in various ways to convince their fellow professionals of the stories the drug companies want told. Partial, after-the-fact disclosure of payments taken for ghostwriting, speaking at conferences etc is not good enough. Bribery is illegal in most business contexts and eveyone understands why. Drug companies should not be in a position to convince the public that drugs are safe, when they are not.
“Drug companies should not be in a position to convince the public that drugs are safe, when they are not.”
Medicine VS drugs who gets to prescribe, and who is ill/who defines illness?
Voluntary drugs, VS involuntary drug use.
Who enforces mental health?
Replace the word “beer” for the word “drug”.
Beer companies should not be in a position to convince the public that beer (ads on TV) is safe, when beer is not safe.
Profit motive destroys everything. Any company’s goal is profit, and regardless of the industry, that drive for profit demands exploitation. The real question is whether business can be married to anything but greed. Seems the answer is no.