Simulating adverse event spontaneous reporting systems as preferential attachment networks: application to the...

Area of Science:

• Pharmacovigilance and drug safety
• Network science applications
• Computational statistics

Background:

• Spontaneous Reporting Systems (SRS) are vital for post-licensure medical product safety evaluation.
• Data mining techniques are used to detect safety signals in SRS databases.
• Existing SRS simulation strategies have limitations in assessing signal detection performance.

Purpose of the Study:

• To develop a novel simulation strategy for SRS databases.
• The strategy is based on plausible mechanisms of database growth over time.
• To provide a more principled and versatile framework for performance evaluation.

Main Methods:

• Developed a simulation strategy using the network principle of preferential attachment.
• Created simulations based on specific SRS databases.
• Used simulations to compare signal detection thresholds for a data mining algorithm.

Main Results:

• Preferential attachment simulations were structurally similar to targeted SRS databases.
• The approach successfully generated signal-free simulations and mimicked known true signals.
• An analysis of the FDA Vaccine Adverse Event Reporting System suggested PRR > 3.0 may offer better signal detection than PRR > 2.0.

Conclusions:

• Network-based SRS simulation using preferential attachment offers a principled and versatile framework for evaluating safety signal detection algorithms.
• Further research should explore diverse simulated signals and data mining approaches.