Signal detection using change point analysis in postmarket surveillance.

Area of Science:

• Pharmacoepidemiology
• Medical Device Surveillance
• Statistical Analysis

Background:

• Postmarket surveillance relies on signal detection for medical product safety.
• Current methods often overlook temporal trends in adverse event data.
• There is a need for methods that incorporate time-series information for more robust signal detection.

Purpose of the Study:

• To apply change point analysis (CPA) for trend analysis in medical product adverse event data.
• To demonstrate CPA's ability to detect shifts in adverse event reporting over time.
• To evaluate CPA as a complementary tool for existing signal detection methods.

Main Methods:

• Change point analysis (CPA) was applied to a spontaneous adverse event reporting system dataset.
• Two CPA approaches were used: detecting changes in mean and changes in variance.
• The methods were illustrated using a neurostimulator adverse event dataset.

Main Results:

• Two significant change points indicating upward trends were identified in June 2008 and May 2011.
• Potential causes for these change points include battery issues and expanded indications for use.
• Periods of unusually low adverse event reports were noted, possibly due to underreporting.

Conclusions:

• CPA can effectively detect changes in the association between medical products and adverse events over time.
• This method serves as a valuable supplement to current FDA signal detection efforts.
• Identifying temporal shifts is crucial for public health regulation, surveillance, and product recalls.