Science for Sale

This has gotten a bit of coverage in the corporate media, but they don't do a particularly good job of explaining, so I'll take a shot. Abstract only for the common folk, I'm afraid, but Dr. Vedula and colleagues, taking advantage of discovery in lawsuits against Pfizer and Parke Davis got documents pertaining to studies they had sponsored on the drug gabapentin (brand name Neurontin). They have been naughty indeed.

As regular readers know, once a drug has FDA approval there is nothing to stop doctors prescribing it for purposes other than the ones for which it is approved. However, drug companies are prohibited from marketing drugs for "off label" purposes. However again, there's nothing to stop them from doing studies of the drug for off-label purposes and getting those studies published, without bothering to go through the more rigorous requirements of actually getting approval for said purposes. However yet again, there are plenty of ways to get studies published in order to create an impression that a drug is useful for some purpose whereas you haven't actually proven any such thing. But you can make money if you make doctors think your drug is useful for all these off-label indications and prescribe it, even though you're just bamboozling them. But who cares about the health and well being of patients when you can make money off of them, right?

So here's the basic technical concept behind cheating. If you do an experiment in which half the people get the drug and half get the placebo, there will be differences between the two groups just because of random chance. So-called statistical significance is a calculation, based on the size of the two groups and the size of an observed difference, of the probability that the difference is due to chance -- a number called p. If it's less than 5%, by convention, we accept that there's a real difference. Yes, that's arbitrary, and there are better ways to assess whether a drug is really useful, but that's the basic starting point.

This only works, however, if you specify exactly what effect you are looking for ahead of time. If you just go ahead after the fact and make a bunch of comparisons between the two groups, chances are something will be different, but the p value for that difference will be bogus. You were just fishing, you happened to find something that is probably purely coincidental, but you can always lie and say that's what you were looking for all along.

Well, that's what Pfizer and Parke Davis did. Actually they did even worse than that. Gabapentin is approved to prevent seizures, but that's not a big market, so the drug companies wanted to be able to sell it for more common problems, specifically migraines, bipolar disorder, neuropathic pain (pain from nerve damage), and nociceptive pain, i.e. what in the vernacular you might call "real" pain, from damaged tissue. So they did a bunch of studies. However, if the study did not find the drug was effective for the primary, pre-specified endpoint, they tried to find a difference in the positive direction for some other end-point. If they succeeded, they lied and said that was what they were looking for all along, and published the study. If they couldn't come up with something, they just didn't publish the study at all.

That way, they hoped to bamboozle doctors into prescribing the drug to people for whom it probably wouldn't work. They have had to pay fines because they were caught doing off-label marketing in this case, but those fines are just a cost of doing business. As Dr. Vedula et al tell us, this stuff undoubtedly goes on all the time but we're just lucky enough to have caught them at it in this one case. Since they aren't trying to get FDA approval, they don't have to register their trials in advance, make any raw data available, or make unpublished studies public. So they can cheat all they want. And they do.

Scum of the earth.