Sequential Safety Monitoring of Rofecoxib for Cardiovascular Adverse Effects

Background: In 2004, a placebo-controlled trial, Adenomatous Polyp Prevention on Vioxx (APPROVe), was halted due to an increased cardiovascular (CV) risk in patients using rofecoxib. Based on this, Merck withdrew rofecoxib from the market in September 2004. However, an estimated number of 80 million people had already taken rofecoxib by the time of its withdrawal.

Objectives: In this methodological study, we performed sequential monitoring of rofecoxib use and CV outcomes using a case-time-control design to investigate how early the association could have been identified if such a monitoring setup were in place.

Methods: Using the Danish population-based health registers, we conducted a case-time-control study to examine the association of rofecoxib with myocardial infarction (MI), ischemic stroke and all-cause mortality during the study period from January 2001 to September 2004. We performed sequential analyses adding 6 months of data from each monitoring period. Incident cases of MI and ischemic stroke were identified using discharge diagnoses registered in the Danish National Patient Register. Exposure status on the date of diagnosis was assessed using a 60-day window. For reference, exposure status was also assessed 120, 180 and 240 days prior to the diagnosis. We matched up to 5 time-controls to each case based on age, sex and calendar time. Exposure odds ratios (ORs) with 95% confidence intervals (CIs) for each outcome were estimated by comparing the odds of exposure during the focal window to reference windows using conditional logistic regression.

Results: We identified 1208 cases of MI, 674 cases of ischemic stroke and 10,243 fatal cases from January 2001 until September 2004. There were 6026, 3370 and 50908 matched time-controls for MI, stroke and all-cause mortality, respectively. The study identified associations of rofecoxib with MI (cumulative OR 1.20; CI 1.05 - 1.37) and all-cause mortality (cumulative OR 1.39; CI 1.33 - 1.45); however, it failed to identify any association with ischemic stroke (cumulative OR 0.98; CI 0.82 - 1.17). For MI, the early estimates from study period 1 (OR 1.06; CI 0.63 - 1.77) and 2 (OR 1.37; CI 0.99 - 1.88) were least precise, but the precision improved gradually. We identified our first signal in study period 3 (OR 1.37; CI 1.02 - 1.67) with the subsequent study periods verifying the signal (OR 1.35; CI 1.10 - 1.65). Later, study periods 5-10 showed a stable OR (around 1.20) with improved precision.

Conclusions: Our monitoring study successfully identified the associations of rofecoxib with higher risk of MI and all-cause mortality. Our findings suggest that the risk of MI could have been established as early as after 2 years of rofecoxib launch using this monitoring setup.