False discovery rate estimation for frequentist pharmacovigilance signal detection methods

Area of Science:

• Pharmacovigilance and Drug Safety
• Biostatistics
• Computational Epidemiology

Background:

• Pharmacovigilance systems detect adverse drug effects using spontaneous reporting databases.
• Current automatic signaling methods rely on arbitrary thresholds, limiting their precision.
• There is a need for more robust signal-generation procedures in pharmacovigilance.

Purpose of the Study:

• To propose a new signal-generation procedure for pharmacovigilance.
• To define a decision criterion based on a critical region for P-values.
• To address the limitations of arbitrary thresholds in existing methods.

Main Methods:

• The new procedure uses a critical region for P-values from the reporting odds ratio and Fisher's exact tests.
• False discovery rate (FDR) is employed to define the critical region.
• P-values mixture model procedures are adapted for one-sided tests and FDR estimation, illustrated with the location-based estimator.

Main Results:

• A simulation study evaluated the proposed methodology.
• The method was applied to the French pharmacovigilance database.
• The approach demonstrated potential for more reliable signal generation.

Conclusions:

• The proposed signal-generation procedure offers a statistically rigorous alternative to arbitrary thresholds.
• Utilizing false discovery rates enhances the reliability of detecting adverse drug events.
• This method can improve the early detection of drug safety signals in large databases.