Disproportionality Analysis and Causal Inference in Drug Safety.

Area of Science:

• Pharmacovigilance
• Causal Inference
• Biostatistics

Background:

• Disproportionality analysis is a common method for detecting drug safety signals from spontaneous adverse event reports.
• Findings from disproportionality analysis are frequently published as real-world evidence regarding drug safety.

Purpose of the Study:

• To review the theoretical properties of disproportionality analysis within the framework of causal inference.
• To identify the conditions under which disproportionality measures can approximate causal rate ratios for drug-event associations.

Main Methods:

• Theoretical review of disproportionality analysis.
• Application of causal inference principles to assess disproportionality measures.
• Analysis of conditions required for disproportionality to approximate causal rate ratios.

Main Results:

• Disproportionality measures can approximate the causal rate ratio only if specific conditions are met: no uncontrolled confounding, uniform or absent under-reporting, and equal overall reporting rates across drugs.
• These stringent conditions are typically not satisfied in real-world data.
• The majority of disproportionality 'hits' likely represent statistical noise rather than true causal associations.

Conclusions:

• Disproportionality analysis, while widely used, often identifies statistical noise instead of causal drug-event associations due to unmet theoretical assumptions.
• Researchers should clearly state the assumptions and exploratory nature of disproportionality findings when publishing.
• Further research may be needed to refine methods for robust drug safety signal detection.