Tuesday, March 20, 2007

Corcept Spins Out

The interesting thing about Corlux (mifepristone/RU-486) is that no matter how it fares in clinical trials, it is always a winner. In the latest trial, Corlux was again not effective on the primary outcome measure, which assessed the psychotic symptoms of psychotic depression. This is not surprising, since it has generally shown mediocre results, which are then spun by the company executives/academics as evidence of treatment effectiveness. Oh, and despite this being pushed as a treatment for psychotic depression, the treatment has never yielded anything resembling efficacy for depression, which strikes me as pretty odd.

Dr. Joseph Belanoff, Corcept CEO, had the following to say about the latest trial results:

While we are disappointed that the trial did not meet the primary endpoint, we are particularly encouraged to have met the important predefined threshold concentration endpoint with statistical significance," said Joseph K. Belanoff, M.D., Corcept's Chief Executive Officer. "This study confirms our previous observation that at higher plasma levels the drug candidate is able to demonstrate desired clinical effects. In particular, those patients in Study 06 who achieved a predetermined level of 1661 nanograms of CORLUX per milliliter of plasma separated from the placebo group with statistical significance.

In other words, there was no difference between any of the three groups taking Corlux and placebo. None. So it appears that they started data dredging (e.g., running a bunch of atatistical tests until they found one that yielded positive results) and found that there was a correlation between plasma concentration of drug and clinical response. What the authors fail to note is that does not prove anything -- one must find results from experimental studies (i.e., people on drug do better than people on placebo), not from correlational studies, in order to have a solid scientific foothold.

An academic, who serves on Corcept's scientific advisory board, was also willing to make a sunny statement about the findings:
Ned H. Kalin, M.D., Hedberg Professor and Chair of the Department of Psychiatry at the University of Wisconsin, said, "The correlation between plasma levels of drug and response rates found in this trial is very exciting. The results of this study show that when psychotically depressed patients achieve a threshold concentration of CORLUX in their system, a rapid and sustained clinical response is likely. This is a strong demonstration of a drug effect in an illness that is potentially devastating and difficult to treat."
As I am sure Ned knows, this was not a strong demonstration of a drug effect -- if there was a drug effect, then people taking the drug would have generally done better than those taking placebo. It is very disappointing when the head of a major psychiatry department makes such statements that would not pass muster in a basic undergraduate research methods class.

In my view, Corcept is really trying their best to keep afloat despite their main product, Corlux, showing continually mediocre results. Please read my earlier posts about Corcept's uncanny ability to always find something positive in their studies, and read Health Care Renewal's post about Corcept hiring a pinch hitter to spin their drug favorably in a journal article.

Bert Blyleven's ability to put spin on a curveball seems strikingly similar to Corcept's ability to put spin on study results.


Anonymous said...

I agree that Corcept, like most pharmaceutical companies, spins lackluster "findings" to make their drug look effective.
But, just to play devil's advocate, if they did unexpectedly find in this study that Corlux had a clinical effect that was related to the drug's blood level (akin to lithium, for example), it doesn't seem unreasonable to mention this. I would assume the next step would be to do a study where they titrated patients to what they consider to be the effective blood level (1661ng) and then compare the results in these patients to those on placebo.

CL Psych said...

If this was the first study they had done on the topic, then I could see doing a little digging to examine correlations. However, when trial after trial fails to yield any notable improvement relative to a placebo, that's when I become highly cynical about this supposedly important correlation between plasma level and outcome.

They're going to do yet another study and we'll wait and see if higher doses lead to improvement.

soulful sepulcher said...

Off label use for mifepristone/RU-486 is going to keep this alive and well on the market.

I've written a bit about it in my series questioning 'hormones or mental illness' --a endo doc needed or psych needed, on my blog.

Anonymous said...

I think the drug must work. A set of results like these should normally leave a company and its researchers thoroughly depressed, and yet their disposition is sunny and optimistic.

What more proof do you need of efficacy?

Anonymous said...

Agree with cl psych about the data dredging.

If you do a enough correlations you're bound to come up with a few significant looking ones by chance.

And if you've got a trial that generates, say, twenty measurable outputs, there's 200 pairwise correlations to play with.

CL Psych said...

Good point Giles!

Anonymous said...

Sorry guys but there has been both research and clinical trials done in the UK on this as well (no relationship to Corcept) and the findings indicate that those taking it did better than others on depression scores etc.

CL Psych said...


If this is true, please provide a way for readers to track down this research. I don't think such research actually exists.