While the protections for individuality make America the colorful place it is today, it sure is easy to run massive cohort studies in European countries where they sacrifice a little bit of anonymity for the common good.
Everyone in Denmark has a number, and they tracked every patient in Denmark with a history of MI to see if they had any adverse events after receiving a prescription for NSAIDs. There were a few significant differences in the populations receiving each different kind of NSAID – rofecoxib and celecoxib tended to be given to older, female populations, and there were some differences throughout their groups regarding the prevalence of other co-administered cardiac medications.
This article really annoys me because the page with which they present their incidence of death by week has six charts that lend themselves immediately to visual comparison – but their chart scales are grossly different. Ibuprofen looks terrible at first glance, but then you realize it has the smallest y-axis scale, and actually performs quite well. In the end, they all demonstrated worsening of outcomes regarding death/MI compared to the total study population rate of death/MI not proximate to NSAID use.
In the end, ibuprofen and naproxen had the least effect on the OR for death; it is fair to avoid rofecoxib, celecoxib, and diclofenac in your routine prescribing without specific indications.