Subscribe Subscribe   Find us on Twitter Follow POL on Twitter  



Statistical significance and Vioxx

In the second Vioxx trial, Humeston v. Merck, the plaintiffs' attorneys are being allowed to examine a Merck scientist and obtain a concession that Merck did not tell its sales representatives that, in one study, 21 Vioxx patients died, while only 9 people who took a placebo died during the study. In cross-examination, Merck drew out the rest of the story: those deaths included such causes such as electrocution and head injury, and the difference was statistically insignificant. (John Curran, AP/SF Chronicle, Sep. 27).

Perhaps the jury will come to the right conclusion, and reject the unscientific proposition the plaintiffs are asking them to make. But why is the argument even in front of the jury? If it isn't per se legal to limit disclosures to statistically significant data, companies can be punished for failing to bury their sales representatives in mounds of irrelevant information. But were Merck, or any corporation, to do precisely what is implied by the plaintiffs' second-guessing, they could be blamed in court for trying to "hide" the important information.

Indeed, in the Ernst case, we saw both tactics at play. Mark Lanier was allowed to pull a "Miracle on 34th Street"-style grandstanding of piling up hundreds of boxes of the paper version of records Merck submitted to the FDA in electronically searchable form, and falsely imply that Merck chose to submit that much paperwork to hide the (statistically insignificant) cardiological results—never mind that all of the data Merck submitted was required by FDA regulations.



Rafael Mangual
Project Manager,
Legal Policy

Manhattan Institute


Published by the Manhattan Institute

The Manhattan Insitute's Center for Legal Policy.