A simulation-based comparison of drug-drug interaction signal detection methods

Area of Science:

• Pharmacovigilance
• Biostatistics
• Drug Safety

Background:

• Post-market drug safety surveillance relies on spontaneous reporting systems to detect adverse drug reactions from drug-drug interactions (DDIs).
• Various statistical methods exist for DDI signal detection, but their comparative performance is not well-established.

Purpose of the Study:

• To assess and compare the performance of different statistical methods for detecting drug-drug interactions (DDIs).
• To identify the most sensitive and specific methods for DDI signal detection in pharmacovigilance.

Main Methods:

• Reviewed six DDI signal detection methods: Ω shrinkage measure, chi-square statistic, proportional reporting ratio, concomitant signal score, additive model, and multiplicative model.
• Conducted simulation studies under various scenarios to evaluate method performance.
• Applied selected methods to the Korea Adverse Event Reporting System (KAERS) database.

Main Results:

• The Ω shrinkage measure and chi-square statistic (threshold=2) demonstrated higher sensitivity in detecting true DDI signals across most simulated scenarios.
• These two methods maintained a false positive rate below 0.05.
• Application to KAERS data identified a known DDI (QT prolongation) and a suspected DDI (hyperkalemia).

Conclusions:

• The effectiveness of DDI signal detection methods can vary.
• Using multiple methods concurrently is recommended for robust DDI signal detection and decision-making in pharmacovigilance.