Lessons from the withdrawal of rofecoxib

Editor—In their editorial Dieppe et al describe the lessons from the withdrawal of rofecoxib.¹

Randomised controlled trials are not the usual way in which serious, uncommon adverse effects of a new drug are discovered. They are usually discovered in observational studies, often case-control studies. Using randomised controlled trials to find adverse effects is dangerous. It would feed on the current vogue from trialists that the only data that count are those from randomised controlled trials. 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 rofecoxib story, the cardiovascular effect is strong enough to see in a randomised controlled trial in 2000, although according to some creative interpretation of the data a protective effect of control drug could have explained the result. The current unpublished randomised controlled trial confirms the adverse effect some four years later.

What if, rather than conducting a randomised controlled trial 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 rofecoxib rather than waiting for results of a randomised controlled trial not designed to find the adverse effect? My hunch is that the case-control study would have implicated the drug sooner than waiting for the randomised controlled trial did. At less cost than a randomised controlled trial, the case-control study would probably have been powered to determine an increased risk that would be missed in a randomised controlled trial, simply because one could not afford a large enough sample in the randomised controlled trial.

Observational studies have value. They have played and they probably will continue to play an important part in the protection of the public health.