BMJ  2004;329:E333 (11 December), doi:10.1136/bmj.329.7479.E333

BMJ USA: Letter

RAPID RESPONSES FROM BMJ.COM

Randomized controlled trials are not the usual way serious, uncommon adverse effects of a new drug are discovered. They are usually discovered in observational studies, often case-control studies. I think there is danger in using RCTs to find adverse effects. It would feed on the current vogue from trialists that the only data that count are those from RCTs. By this standard, we would not have concluded that cigarette smoking causes lung cancer. We should not exclude observational studies, but rather continue to do them well in a timely fashion.

As I understand the Vioxx story, the cardiovascular effect is strong enough to see in an RCT in 2000, although there was some creative interpretation of the data suggesting that a protective effect of control drug could have explained the result. The current unpublished RCT confirms the adverse effect some 4 years later.

What if, rather than conducting an RCT on a few thousand more patients, a large case-control study had been conducted to test the hypothesis of whether or not the increase was from Vioxx rather than waiting for results of an RCT not designed to find the adverse effect? My hunch is that the case-control study would have nailed the drug sooner than waiting for the RCT. At less cost than an RCT, the case-control study would likely have been powered to determine increased risk that would be missed in an RCT simply because one could not afford a large enough sample in the RCT.

There is value in observational studies. They have played and they likely will continue to play an important role in the protection of the public health.

Rollins School of Public Health of Emory University 1518 Clifton Rd NE, Atlanta, GA 30322

Competing interests: Co-inventor on patent that covers putting folic acid in oral contraceptives. A paid consultant to Johnson & Johnson on this issue.

CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    Technorati    What's this?