Unbalanced reporting of benefits and harms in abstracts on rofecoxib

It was predicted from the mechanism of action that, compared to older non-steroidal anti-inflammatory drugs, rofecoxib (Vioxx) would reduce gastrointestinal bleeding, but also that it would increase the occurrence of cardiovascular thrombosis. From the patient's point of view, both effects are important and should be investigated and reported similarly. We studied how they have been reported over time.

We searched PubMed for abstracts on rofecoxib that commented on gastrointestinal bleeding or cardiovascular thrombosis or both. Two researchers, blinded to date of publication and authors, assessed the abstracts independently. We judged the authors' view on rofecoxib and comments on gastrointestinal bleeding and thrombosis as being favourable, neutral or unfavourable towards rofecoxib.

We included 393 abstracts commenting on gastrointestinal bleeding (72%) and cardiovascular thrombosis (54%) or both. Before October 2000, all abstracts (n = 27) mentioned only gastrointestinal bleeding and 89% were positive towards rofecoxib. The year before the withdrawal of rofecoxib (October 2003 to September 2004) (n = 46), 59% of abstracts commented on gastrointestinal bleeding only, 17% on thrombosis only, 24% on both and 67% were still positive. From October 2006 to September 2007 (n = 54), 13% mentioned gastrointestinal bleeding, 54% thrombosis, 33% mentioned both and only 11% were positive.

The reporting of benefits and harms was not balanced and changed markedly over time. Knowledge of increased risk of thrombosis existed early on, but the harms came into focus too late, when the drug was already withdrawn, and when tens of thousands of patients had been harmed unnecessarily.