Sequential Epidemiological Analyses of Real-World Data: A Tool for Prospective Drug Safety Surveillance from the Rofecoxib Example

Introduction Large administrative healthcare databases can be used for near real-time sequential safety surveillance of drugs as an alternative approach to traditional reporting-based pharmacovigilance. The study aims to build and empirically test a prospective drug safety monitoring setup and perform a sequential safety monitoring of rofecoxib use and risk of cardiovascular outcomes. Methods We used Danish population-based health registers and performed sequential analysis of rofecoxib use and cardiovascular outcomes using case–time–control and cohort study designs from January 2000 to September 2004. Each monitoring period added 6 months of data until the end of the study period. In the case–time–control study, incident cases of myocardial infarction (MI) and ischemic stroke were identified and matched with up to five time controls on age, sex, and calendar time. Exposure status on the date of diagnosis was assessed using a 60-day focal window, with reference windows 120, 180, and 240 days prior to the diagnoses. In the cohort study, incident users of rofecoxib were matched up to 1:4 with ibuprofen users (active comparators) using high-dimensional disease risk scores and were followed for 60 days. Results The earliest association between rofecoxib use and the risk of MI was seen in study period 2 for case–time–control design (OR 1.42, 95% CI 1.04–1.93) and in study period 7 for the cohort study design (RR 1.22; 95% CI 1.02–1.47). Conclusions Our prospective drug safety monitoring setup showed that the risk of MI could have been detected 3.5 years before the ultimate market withdrawal of rofecoxib. However, further research is needed to validate this approach.