Using Data Mining to Predict Safety Actions from FDA Adverse Event Reporting System Data
- PDF / 161,149 Bytes
- 11 Pages / 612 x 792 pts (letter) Page_size
- 64 Downloads / 206 Views
633
Using Data Mining to Predict Safety Actions From FDA Adverse Event Reporting System Data
Alan M. Hochberg Vice President, Research Stephanie J. Reisinger Senior Vice President, Research Development Ronald K. Pearson, PhD Senior Scientist Donald J. O’Hara Senior Software Engineer Kevin Hall, MD Consultant ProSanos Corporation, Harrisburg, Pennsylvania
Key Words Pharmacovigilance; Data mining; Drug safety; Adverse event reporting system; Algorithms Correspondence Address Alan Hochberg, ProSanos Corporation, 225 Market Street, Suite 502, Harrisburg, PA 17101 (e-mail: alan.hochberg@ prosanos.com).
Purpose: To determine the value of data mining in early identification of drug safety signals from spontaneous reporting databases. Methods: A single data mining algorithm was applied to the 2001–2003 public release of Food and Drug Administration Adverse Event Reporting System (AERS) data for all therapeutic new molecular entities (NMEs) approved in 2001. The list of detected signals was compared with the list of safety-related regulatory actions for those drugs through February 2006. Results: For the 21 NMEs, 73 signals of interest were detected by data mining. In 39 cases, that signal preceded regulatory action.
INTRODUCTION Postmarket surveillance for adverse events is an essential component of every national and regional system for assuring drug safety. Systems such as the Food and Drug Administration Adverse Event Reporting System (AERS) were primarily designed to capture rare, serious events that were not recognized in clinical trials because of a low frequency of occurrence (1). Such a system is most effective when the background rate of the adverse event is relatively low, such as in the case of drug reactions leading to events such as rhabdomyolysis or Stevens-Johnson syndrome. One pharmacovigilance investigator has characterized this as the “worst first list”: serious adverse events that almost invariably attract the attention of safety investigators for case-by-case scrutiny and thorough discussion of their implications. Increasingly, attention has been focused on the use of AERS data in pharmacovigilance beyond the worst first list (2). This includes the use of data mining algorithms to detect drug-related increases in the rates of relatively common medical conditions. The need to detect such increases has been highlighted by safety issues
The median time from approval to signal detection was 11.5 months, and the median time from signal detection to action was 21 months. There were 33 actions for which no signal was detected and 34 signals with no corresponding regulatory action. Conclusion: Using AERS data 2–3 years following approval, more than half of FDA actions that occurred in the next 2–4 years were predicted by data mining, and more than half of the signals detected by data mining corresponded to an FDA action. An appropriate data mining procedure can yield meaningful safety information, often well in advance of regulatory action.
such as cardiac events as a consequence of COX-2 inhibitor use (
Data Loading...
