Nonsteroidal anti-inflammatory drugs (NSAIDs) are among the most frequently prescribed medications in the United States. Because efficacy of both cyclooxygenase (COX)-2 selective and traditional NSAIDs is largely equivalent, treatment decisions about NSAIDs hinge on their comparative safety. The Medicare Current Beneficiary Survey (MCBS), a longitudinal data source that links Medicare claims data with in-home surveys, provides a unique amalgam of both administrative and patient-derived data that can be used to address NSAID ? questions. Use of a novel methodology based on DuMouchel's Multi-item Gamma Poisson Shrinker (MGPS) algorithm offers potential advantages over traditional analytic methods including the ability to detect effectiveness and safety signals in a more timely and less-resource intense fashion. The MGPS has been mostly studied using the FDA Adverse Event Reporting (AERS) database and has had limited application to other datasources. As a proof-of-concept study, we propose to use MCBS to test important hypotheses regarding NSAID-related safety. ? ? The Specific Aims for this project are to: ? 1. Adapt survey data from the Medicare Current Beneficiary Survey (MCBS) linked to Medicare claims data for use with both cohort analyses and Bayesian data-mining algorithms as a model for an ongoing data source useful for post-marketing studies of the comparative effectiveness and safety of drugs. ? 2. Conduct a retrospective cohort analysis using MCBS with Medicare linkage to: 2a) Examine the risk for acute myocardial infarction among coxib users compared to traditional NSAID users and NSAID non-users. 2b) Examine the risk for stroke among coxib users compared to traditional NSAID users and NSAID non-users. 2c) Examine the risk for heart failure among coxib users compared to traditional NSAID users and NSAID nonusers. 2d) Examine the effectiveness of coxibs in reducing the risk of Gl bleeding compared to traditional NSAID users and non-NSAID users. ? 3. Use the Bayesian-based datamining approach (i.e. the MGPS algorithm) to identify NSAID-related adverse outcomes within MCBS with Medicare linkage and compare with Specific Aim 2 results derived using cohort analyses. Following this initial modest investment in solving methodologic challenges inherent in using MGPS with longitudinal data, MPGS techniques will be generalizable to Medicare Part D data, and other drugs and disease states. With the combined expertise of UAB CERTS investigators facile with the use of Medicare administrative data and FDA researchers and statistical industry consultants experienced in using MGPS, the project proposed has significant promise to provide new methods for pharmacovigilance of national public health importance. ? . ? ?
Showing the most recent 10 out of 32 publications
