StATS: Data mining and drug safety (created 2006-05-04).

I am very interested in safety issues, especially in the continuing review/interim analysis of clinical trials. It turns out that S-plus is targeting drug safety as a particularly important application of its data mining modules. Two recent web seminars addressed this topic:

• Pre-Analytic and Post-Analytic Factors in Post market Drug Safety Data Mining. Speakers: Alan Hochberg of ProSanos Corporation and Michael O'Connell of Insightful Corporation. Abstract: The pharmaceutical industry is now responding to a series of high-profile drug safety issues in a number of ways, including the development and deployment of new methods of visualizing and mining drug safety information. Many of these methods employ state-of-the-art techniques in statistical analysis and graphical display. In developing and deploying these methods, it must be kept in mind that pharmacovigilance is a man-machine partnership: a computer is used to collect, digest, and display large amounts of data regarding drugs and adverse events, while a human drug safety expert, trained in biochemistry, physiology, and medicine as well as statistics, interprets the data and makes decisions.
• Safety Data Analysis and Reporting: Signal Detection, Data-Mining and Next Generation Methodology for Drug Risk Assessment and Safety Research. Speakers: Alejandro Murua, University of Montreal and Michael O'Connell, Insightful Corporation. In its efforts to continue to serve the public health and protect public safety, the pharmaceutical industry is now challenged to set up sound pharmacovigilance plans that carefully analyze and report on pre-marketing clinical study data, minimize risk and monitor post-marketing safety. These plans include the statistical analysis and reporting of adverse events in clinical studies and the proactive analysis of observational data such as FDA AERS for signal detection and risk management. In this context, the use of sophisticated statistical analysis models and data mining techniques is of increasing importance in pharmacovigilance, as these methods are able to detect signals earlier and more accurately than current methods.

There are not any public links to these presentations, but you should be able to register at the S-plus website (www.insightful.com) and get access to them.

I want to use control charts to monitor safety data. I think it is a nice complement to the more complex data mining approaches advocated here (and by others as well). The control charts are a simple approach that pretty much anyone can use. It helps distinguish between signals and noise and has been optimized over years of work in business and industry. Data mining tools can look at more complex interrelationships among variables, and will be more effective in the hands of a trained statistician that a control chart would. Nevertheless, a control chart is still valuable because it will be used more often by non-statisticians and will allow more people to engage themselves in the examination of safety issues.

This page was written by Steve Simon while working at Children's Mercy Hospital. Although I do not hold the copyright for this material, I am reproducing it here as a service, as it is no longer available on the Children's Mercy Hospital website. Need more information? I have a page with general help resources. You can also browse for pages similar to this one at Category: Adverse events in clinical trials.