Friday, July 17, 2009

A Further Retrospective on Xigris: Approved Against the Rules?

Poole and colleagues recently completed a retrospective analysis of the approval process for Xigris (drotrecogin alfa) in both the US and Europe, that raises some further issues about this drug.

In HOOKED, the story I told about Xigris mostly postdated the time of approval. I stated that this much-heralded drug for sepsis turned out to be quite disappointing, offering relatively low rates of success in the face of extremely high costs. The result was that many hospital ICUs balked at the price tag. The manufacturer, Eli Lilly, responded with a bold PR counterattack, claiming that any hospital that withheld Xigris from its patients was engaging in unethical bedside rationing of care. They also hired on Sandra Tiffany, a Nevada state senator, who believed that her life had been saved by Xigris and then gave numerous TV interviews.

Poole and colleagues now expand the account by looking more carefully at the approval process both in the US, at the FDA, and in the counterpart European agency (EMEA). They found that both agencies broke their own rules by approving Xigris based on only one randomized trial. The trial that documented the supposed benefits of Xigris, called PROWESS, was as noted disappointing from the manufacturer's viewpoint. On the one hand it showed a statistically significant improvement of drug vs. placebo. On the other hand the effect size in favor of the drug was nothing to write home about. That apparently led the investigators to perform a large number of posthoc subgroup analyses. They found one such subgroup, those with severe sepsis according to APACHE III score, that seemed to do better on Xigris than the other subgroups.

It would seem as if the FDA and the EMEA, having violated rule #1 by approving a drug based on a single study, proceeded to try to make up for it by violating rule #2--don't approve a drug based on a posthoc subgroup analysis. That is, having approved Xigris, they then put a label restriction on it, that it was only for use in severe sepsis. The folly of that move, according to Poole et al., is that later studies revealed that the beneficial results in that subgroup could not be replicated and probably were a statistical fluke--which is precisely why we are warned by statisticians not to trust posthoc subgroup analyses, commonly poo-poohed as "data dredging."

Against these critical views of the drug approval process go our understanding of the pressure that was undoubtedly felt by the agencies to approve this drug, for two reasons--first, that sepsis has no effective treatment and is a high-mortality condition; and second, that Xigris is something of a breakthrough product that replies on a new chemical mechanism of action. Today we can say that it has proved much less effective than was initially thought, and in addition seems to cause a serious amount of bleeding as its major adverse reaction. It can probably said with justice, even in hindsight, that if the FDA and the EMEA were going to approve the drug contrary to their own rules of best evidence, they might at least have done so with the mandate that the company perform more extensive follow-up studies. As it was, Lilly was left totally on its own as to whether or not to do those studies.

Thanks incidentally to Rick Bukata and Jerry Hoffman of Primary Care Medical Abstracts for calling my attenton to this paper.

Poole D, Bertolini G, Garattini S. Errors in the approval process and post-marketing evaluation of drotrecogin alfa (activated) for the treatment of severe sepsis. Lancet Infectious Diseases 9:67-72, 2009.

2 comments:

Dan said...

Great post, doc. I interviewed to sell this useless drug years ago, and asked questions about it's efficacy to Eli Lilly during these interviews. They did not hire me.

Unknown said...

The other lesson is that if you dig hard enough in either study data or pharmacoeconomic patterns, you can find a sliver of justification for anything.

I suspect that the development team was in love with the idea of how Xigris was supposed to work, and had just thrown too much effort and $ into the drug to back down when the study results didn't support that core idea.

Disclaimer: I once worked with people who crunched numbers for the XEUS and Mercury trials that examined Xigris data