Calculating Expected Health Losses in Prospective Drug Safety Monitoring

• Activity Number: 82
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Jessica Myers*+ and Amanda Patrick and Milton Weinstein and Robert Glynn and Sebastian Schneeweiss
• Companies: Brigham and Women's Hospital and Brigham and Women's Hospital and Harvard School of Public Health and Harvard Medical School and Brigham and Women's Hospital
• Keywords: Bayesian decision theory ; group sequential methods ; observational studies

Abstract

The Food and Drug Administration's Sentinel Initiative aims to use existing electronic healthcare data for prospective safety monitoring of regulated medications and to facilitate the assessment of the impact of FDA regulatory actions on public health. A key issue is determining when evidence about a potential safety issue is sufficient to warrant regulatory action. In this paper, we utilize Bayesian decision theory to provide a framework for calculating expected health losses (weighted event counts for multiple outcomes) in the population under several potential FDA strategies, including removing a drug from the market or further observational monitoring. This framework has the advantage that assumptions that inform the decision process must be quantified and clearly specified. Our calculations may be seen as a solution to a simplified two-armed bandit problem. We apply our framework to monitoring the safety and efficacy of the recently approved drug prasugrel as compared to clopidogrel, a drug with similar indications, in sequential propensity-score matched cohorts.