« Mid-day Update |
| Can't Buy These Thrills »
May 22, 2006
Merck and the Numbers
The New York Times has a good article today on the Vioxx data that I was talking about here last week. Check the graphic of the Kaplan-Meier charts especially; it's a good illustration of the problem. Merck is technically correct that the latest data still don't show a statistically meaningful difference between the Vioxx group and placebo until at least 18 months. As the article makes clear, they're hitting that theme very hard.
But Merck is also living in a dream world if they think that's going to help them much at this point. The problem is, the data look as if they're trending worse from a much earlier stage, and finally reach significance at the later time points. No lawyer in the world is going to walk away from that without driving it into the jury's heads that the danger is plain to see, yes, right there from the beginning, and don't talk to me about p-values when anyone can just look at this chart - your chart! - and see what's really going on. . .etc. We live by statistical arguments in the drug industry, but the people who are being called to jury duty sure don't. If I were one of the plaintiff's attorneys, I'd use the voir dire to make sure that anyone who knew anything about statistics never saw the inside of the jury box.
What's worse, to nonscientists, making statistics the centerpiece of your defense sounds shifty. People don't trust them; it's not for nothing that there are all those variously attributed quotations about "Lies, damned lies, and statistics". Now, if someone asks "Why are you so sure?" about something where I work, the answer "p less than point-oh-oh-five" will stop the questioner in their tracks. Not so in most workplaces, where that answer would make you sound as if you're dodging the question. And let's face it, the only p-values that strong that Merck can show are the ones that work against them.
The other problem is that a statistical approach is valid for large samples, the larger the better. But the jury isn't looking at a large sample. They're not there to decide how much Vioxx might have raised aggregate cardiovascular risk in certain subgroups, they're there to decide if it caused a heart attack in that guy sitting over there. The attorneys are going to keep things as personal as possible.
+ TrackBacks (0) | Category: Cardiovascular Disease
POST A COMMENT
- RELATED ENTRIES
- Molecular Printing of Drug Molecules. Say What?
- Pfizer Walks Again By Night
- Gitcher SF5 Groups Right Here
- Changing A Broken Science System
- One and Done
- The Latest Protein-Protein Compounds
- Professor Fukuyama's Solvent Peaks
- Novartis Gets Out of RNAi