Gavelwrench Fishfire | A Blog

Hey folks, time for the second in my 75 part biennial Boring Lecture series.  Last time I discussed how more freedom in the market doesn’t necessarily lead to more efficient markets, even with major assumptions.  Today I’m gonna talk about something else that’s been mentioned in the news (a month ago).

This summer I’ve been immersed in biostatistics due to the nifty little summer program for undergraduates here at THE University of Wisconsin.  I have been helping out with questions the undergraduates have about the statistical program we use here, and attending some of their lectures about clinical trials.  Right after returning from one of these lectures, Julia mentioned this blog post to me, which criticizes the FDA for failing to approve a new drug named Flibanserin to treat “hypoactive sexual desire” in women.  For those of you who, like me, are not good with prefixes, “hypo-” means below normal.  People who believe that the rejection of this drug was due to some sort of sexism that readily pushes Viagra but fails to acknowledge the possibility of a real disorder in women are ignoring the basic way the FDA and statistics work.

The previously mentioned blog post takes the view that the side effects were not severe enough to reject the drug, especially compared to the side effects of Viagra.  I think this would be a valid point, and I even think that the FDA doesn’t have any business telling us what kind of side effects we can and can’t accept.  But in this case, the drug showed some side effects, and it may not actually work. That bolded point is where FDA haters usually get it wrong.  When we see that something has been approved by the FDA, we naturally assume that it’s effective.  Therefore the FDA has a responsibility to weed out as many ineffective drugs as possible.

I’ve actually looked at the FDA briefing on the Flibanserin trial, which is here.  The study aimed to show two things:  an increase in sexual desire and an increase in the number of “sexually satisfying events,” or SSEs.  The FDA and the drug company agreed in previous meetings that these two things had to be shown in order to show efficacy.  The trial showed a significant increase in SSEs, but no significant increase in desire (page 30 of the linked document).  So as far as the FDA is concerned, that’s it.  The drug has not been shown to be effective, therefore even in the absence of side effects it can’t be approved.

The drug company behind Flibanserin tried as hard as they could to convince the FDA that their drug worked.  The company petitioned to switch the way desire was measured.  Some other measurement besides the daily logs was suggested, but this is a big no-no from a statistical standpoint.  Probably the most important rule about clinical trials is that you can’t change what you’re testing after the test is over.  This is exactly what the company wanted to do, and if the FDA allowed this kind of thing, we would have many more ineffective and dangerous drugs on the market.

As kind of side note, another trial was done which sought only to show that Flibanserin increases SSEs failed to show an effect.  Although this isn’t actually that relevant, since the drug didn’t pass the efficacy standards agreed upon in the first place.

Of course, I want to say that the blog linked above seems like a fine blog, but the fact of the matter is this clinical trials business is hard.  I think it’s a shame that the FDA gets so much hate, when a lot of it really seems unwarranted.  I’ve heard some stories of heroic statisticians in the FDA standing up to influential statisticians hired by drug companies.  And I don’t even think the statisticians working for the drug companies are doing anything wrong.  They truly believe the drug works, but not all effective drugs can be shown to be effective.  All the parties involved just want to help the most people.