Improving the Efficiency and Rigor of Pharmacovigilance at FDA
CERSI Collaborator: Russ Altman, MD, PhD, Stanford University
FDA Collaborators: Ruth Barratt, PhD, DVM, Robert Ball, MD, MPH, Mitra Rocca, Dipl. Inform. Med.
Carol Pamer, RPh, MPH, Scott Proestel, MD, Steve Anderson, PhD, MPP
Project Start Date: April 2014
Regulatory Science Challenge
FAERS (the Food and Drug Administration Adverse Event Reporting System) and VAERS (Vaccine Adverse Event Reporting System) are used by health care professionals, consumers, and manufacturers to report information on adverse events that may be caused by a drug or vaccine. FDA clinical reviewers evaluate these two data sources manually and use them to track the safety of approved drugs and biological products in the U.S. The number of adverse event case reports that FDA receives exceeds 1 million, and increases annually. Therefore, new, effective, and computer-based methods to monitor drug safety in FAERS and VAERS are needed so that FDA reviewers can focus on the most important reports that affect patient safety.
Project Description & Goals
Researchers will use computer science approaches, such as natural language processing and machine learning to analyze text contained in FAERS and VAERS reports. Researchers will collaborate with medical reviewers in analyzing reports to confirm the system’s accuracy. The goal is to establish an automated system that can quickly and accurately identify safety concerns about marketed biological products reported in VAERS and FAERS so that FDA can further evaluate any issues.
Researchers will also evaluate SIDER2, a publicly available database that contains information on marketed drugs and their side effects. Researchers will compare information on several drugs contained in SIDER2 to information on the actual drug labels. The goal is to evaluate the suitability of SIDER2 as a source of information and record of known side effects. Study findings can be used to improve the performance and accuracy of SIDER2, or other similar databases. In addition, the researchers will assess whether methods used by SIDER2 to extract data could also be used to detect key information in FAERS and VAERS reports.