How Not to Handle Your Clinical Data

We turn now to Orexigen, one of the small companies trying to make headway in the obesity market. Earlier this year, a patent application from them published, claiming that their drug (Contrave, a sustained release formulation combining the known drugs naltrexone and bupropion) had cardiovascular benefits above and beyond its weight-loss effects. Problem was, they based that claim on the first 25% of the data from an ongoing clinical trial. You’re not supposed to go around doing that. The disclosure led to a public fight with researchers at the Cleveland Clinic (including well-known cardiologist Steve Nissen) about whether the trial had been compromised, and whether Orexigen was misleading its shareholders.
Now, as Matthew Herper reports, the trial has been halted, on the grounds that it has indeed been compromised by that premature data release. The FDA had also stated that they would not accept the trial results for that very reason. Orexigen and their partner, Takeda (who must be wondering how they got into this) announced this yesterday morning, followed a half-hour later by the release of the (half-completed) trial data by the Cleveland Clinic team. (Takeda is also talking breach-of-contract). The cardiovascular benefit that had been noted at the quarter-way mark had slipped below statistical significance, and the best bet is that it was on its way to disappearing entirely, had the trial actually completed.
It’s been ugly. Orexigen released a statement, trying to make their case. Read it at that link and see what you think. I have trouble believing, myself, that they’re really pleased that the study is being terminated, and that it’s yielded important information nonetheless.
The company also seems to be trying (obliquely) to blame the USPTO for disclosing the data, but they’re glossing over the fact that (1) everyone knows the schedule for publication of patent applications, and (2) the PTO is only publishing what you put into your own patent application that you wrote yourself. So this is not even as plausible as blaming the “Reply All” button in an e-mail application. They also seem to be blaming Matt Herper for merely reporting the ways in which the company seems determined to super-glue a clown wig onto its head.
I’ll gladly stand by that opinion, in case Orexigen’s people get around to reading this, and I’ll throw in another one: they would have had trouble paying someone to mess up that trial as throughly as they managed to mess it up themselves. For all the talk in the Orexigen statement about all the tensions and problems with disclosure of clinical data, it’s really not all that hard: you shut up about your interim data, unless you’ve already (from the beginning) planned to disclose something, and the trial has been designed with that in mind. Otherwise, unless the numbers are enough to halt the trial (in either a good or bad way), you just sit tight and see what happens. Because interim data are, well, interim. You run a full trial to see what the real numbers are – until then, they aren’t the real numbers.
Update: more on the data release.