To the Editor: Langton et al document the history of rofecoxib approval in 1999 and withdrawal in 2004. 1 They also provide an outline of the possible mechanisms for increased cardiovascular risk, which we detailed in the Journal in August 2001. 2 Their editorial raises the issue of whether the increased cardiovascular risk is a class effect of all selective cyclooxygenase 2 (COX-2) inhibitors and notes that no increased risk has been identified to date with celecoxib use. While this is correct, the editorial omits to state that, although several drugs are categorised as “COX-2 inhibitors”, the selectivity for COX-2 over COX-1 inhibition varies greatly between different drugs (see Box). 3

It is potentially significant that celecoxib is only modestly COX-2 selective compared with rofecoxib. Because COX-2-selective inhibition can lead to selective inhibition of vascular prostacyclin synthesis with little or no effect on vascular or platelet thromboxane synthesis,1 a highly selective COX-2 inhibitor such as rofecoxib is expected to disrupt the balance between antithrombotic prostacyclin and prothrombotic thromboxane. The relatively modest COX-2 selectivity of celecoxib may be one explanation for the lack of adverse cardiovascular effects demonstrated to date. It would also explain its lack of upper gastrointestinal tract protection relative to diclofenac, as both drugs have similar COX-2 selectivity.4

The newer coxibs, like rofecoxib, are significantly more COX-2-selective than celecoxib and, if this selectivity is the basis for the adverse cardiovascular events, then caution is needed with these newer agents. Although the editorial states that trials have not shown increased risk with the newer coxibs, this is not correct. On 15 October 2004, Pfizer announced that valdecoxib, when used for pain management in coronary artery bypass surgery, caused an increased number of adverse cardiovascular events.5

As the editorial states, the VIGOR study with rofecoxib in treating rheumatoid arthritis revealed a greatly increased incidence of adverse cardiovascular events compared with naproxen, and yet rofecoxib sales continued for another four years at a high level. 1 Perhaps the most important question for prescribers arising from the experience with rofecoxib is not whether clinical trial results are conclusive, but how should prescribers respond to apparent conflicts in the medical literature. In such a situation, resort to ethical and legal obligations for disclosure of information will be the prudent approach, as we have detailed.6

For prescribers considering the loss of rofecoxib, some perspective is provided by the following. The number needed to treat (NNT) to cause an increase in one fatal or non-fatal cardiac event in the VIGOR study was 225 (average trial duration was 9 months). In trials with statins in which coronary heart disease was absent at enrolment, the NNT per year to prevent one fatal or non-fatal coronary event was 217 to 256.7