I've saved this "fun with numbers" quote for a long time:
"Is 2.2% different than 0.6%?"
-- Dr. Randall Whitcomb, Warner-Lambert's vice president for diabetes research in 1996, explaining why the company told the FDA that liver injuries for patients taking Rezulin mirrored those in patients taking placebos during clinical trials. In the trials, 2.2% of Rezulin patients suffered liver damage, compared to 0.6% of placebo patients. (LA Times Sunday March 11, 2001)
Now let's fast forward to the CDC's revelation that it may have inflated the number of deaths related to obesity by tens of thousands because of statistical errors. The estimates appeared in a widely-cited article in the Journal of the American Medical Association, stating that obesity is gaining rapidly on tobacco as the leading cause of preventable death. Harvard's George Blackburn is quoted by USA Today saying, "Focusing on this one number is silly."
Is it?
Stanton Glantz of UCSF says in the NY Times, "The kind of policies one would develop for something that is killing about as many people as tobacco or a quarter as many people as tobacco are very different." The Washington Post quotes an analyst for the Center for Consumer Freedom: "A full investigation into the obesity death tally will reveal multiple flaws that seriously overstate the obesity problem and is leading to knee-jerk policymaking and litigation." (Of course, the Center is a "coalition of restaurants, food companies, and consumers working together to promote personal responsibility and protect consumer choices.")
The Wall Street Journal reported that CDC Director Julie Gerberding "conceded that the views of dissenting scientists hadn't been properly heeded" prior to the release of the report.
"I regret that the internal scientific concerns didn't come to light before the paper was published," Dr. Gerberding said. Of the agency's clearance process, she said, "We're going to fix it."
Good. Public policy must be based on statistically-sound data.
Posted by schwitz at November 24, 2004 12:26 PM