Category: Policy and politics

What do MDs need to know about statistics?

published date
September 19th, 2006 by

What do MDs need to know about statistics?

From the Boston Globe (Recommended for MDs: a grounding in statistics)

Doctors need at least a year’s study of statistics to understand the medical advances reported in scientific journals, according to a new report. And most of them don’t have it…

[Most schools provide six to eight weeks of statistics training.]

If doctors can’t interpret information about up-to-date techniques, they might use the information incorrectly or not use it at all, said the report’s author, Nicholas J. Horton, a statistics professor at Smith College…

But not all statisticians think doctors should learn their trade. Stephen Lagakos, a biostatistician at Harvard’s School of Public Health, said most doctors need to know only “the bottom line” of medical studies. He said busy clinicians should be able to use a study’s results without analyzing the statistics behind it.

Most doctors “take it as a given that the journal has done a responsible job in sorting out the bad studies and keeping the good ones,” he said.

I don’t think either of these profs have things quite right. All else being equal, it would be great for docs-to-be to become experts in stats. But expanding the amount of time devoted to statistics, as Horton suggests, would have to come at the expense of something else. I’m not sure it’s worth it.

Lagakos is a little too trusting of medical journals. It’s not just a matter of how well journals sort the good studies from the bad but how those studies are represented by others with a vested interest in getting doctors to act in certain ways, e.g., writing prescriptions for particular drugs.

I’d suggest devoting more of the existing six to eight week stats training to how to be a savvy customer of statistically-based messages. The class could be taught by someone with experience in pharmaceutical or medical device marketing. The syllabus should include titles such as How to Lie with Statistics that get students thinking more critically about what they are hearing.

I’m upset about the Bextra withdrawal

published date
April 8th, 2005 by

A close friend of mine has rheumatoid arthritis. Under the supervision of her rheumatologist she’s tried all the prescription and OTC options and found that Pfizer’s Bextra –and not any other COX-2 inhibitor or other medication– worked well for her. I know someone else with a similar story about Vioxx.

Yesterday the FDA asked Pfizer to withdraw Bextra from the market. A number of cardiovascular, gastro-intestinal, and dermatological adverse events –some fatal, along with Bextra’s failure to prove its superiority to other treatments, doomed the drug. In an unusual move, the FDA overruled its Advisory Panel, which had recommended that the drug be allowed to stay on the market. The only hope the FDA left for patients who feel they need Bextra was to state that a proposal for a compassionate use program would be received favorably, if Pfizer wants to propose one.

Something’s gone terribly wrong here. The pharmaceutical companies made a big mistake by promoting COX-2 inhibitors to the widest possible customer base. As a result it looks like a number of people who should have been treated with the occasional Tylenol or nothing at all ended up sick or dead. And after the FDA was caught asleep at the switch, it may have overreacted to the latest data. As I wrote last month, J&J is taking the lead in balancing risks and benefits in its direct to consumer advertising –this is in the industry’s own best interest.

I read Marcia Angell’s book, The Truth About the Drug Companies when it came out, and there is a lot of truth in there. But I disagree with one of her main arguments –that me-too drugs (similar drugs in the same class) are bad and should not be allowed. Celebrex, Vioxx, Bextra, and the rest may be quite similar, but for whatever reason each seems to work better in some patients than others. There are similar stories in other drug classes, notably in drugs for depression. It doesn’t matter to me whether the motives of the drug companies are pure; I’d rather have more choice than less.

In 20 years or so, when pharmacogenomics and personalized medicine are the norm, we won’t have to go through trial and error treatment with many different drugs. But for now it’s the best we can do.