Design and analysis in a safety investigation of medical products in the Sentinel System often require tailored specifications to the products and outcomes of interest. The goal of this project is to uncover differences in specifications which may be thought of as minor but may greatly impact medical products’ safety analyses. This project has two main phases. The first phase compares the statistical code and specifications for two sets of analyses run in the Sentinel System. These analyses were of special interest because they were thought to have similar design and analysis elements (or specifications) to each other but led to larger-than-expected differences in results on a particular outcome. The second phase of this project runs a multifactorial experiment on a test case. This experiment slightly co-varies multiple specifications at a time and measures their impact on a few safety outcomes. The findings from this project will inform choices and standardization of certain specifications in future drug safety investigations. This will help minimize potential bias and facilitate replication of study findings or design of sensitivity analyses.
Rima Izem, PhD; Office of Biostatistics, Center for Drug Evaluation and Research, U.S. Food and Drug Administration, Silver Spring, MD
Judth C. Maro, PhD, MS; Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, MA
Michael Nguyen, MD; Mark Levenson, PhD; Center for Drug Evaluation and Research, FDA, Silver Spring, MD
Ting-Ying Huang, PhD; Laura Shockro, BA; Laura Hou, MPH, MS; April Duddy, MS; Andrew Petrone, MPH; Ella Pestine, MPH; Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, MA