A Precision Mixture Risk Model to Identify Adverse Drug Events in Subpopulations Using a Case-Crossover Design

Area of Science:

• Pharmacovigilance and Pharmacoepidemiology
• Biostatistics and Health Data Science
• Computational Health and Clinical Informatics

Background:

• Pharmacovigilance studies using real-world data are crucial for detecting adverse drug events (ADEs).
• However, ADE risks in specific subpopulations require enhanced scrutiny to protect vulnerable individuals.
• The case-crossover design, applied to administrative claims data, offers a method for ADE detection while controlling for confounding effects.

Purpose of the Study:

• To propose a novel precision mixture risk model (PMRM) for identifying ADE signals within subpopulations.
• To leverage the case-crossover design for enhanced ADE signal detection in vulnerable patient groups.
• To control for false discovery rate (FDR) and confounding effects in subpopulation-specific ADE signal identification.

Main Methods:

• Implementation of the precision mixture risk model (PMRM) within a case-crossover framework.
• Application of the PMRM to large-scale administrative claims data.
• Analysis of ADE signals across subpopulations defined by demographics, comorbidities, and diagnosis codes.

Main Results:

• The PMRM successfully identified ADE signals in various subpopulations.
• Certain drugs demonstrated ADE risks exclusively in subpopulations, not in the general population.
• The PMRM controlled FDR effectively and showed higher sensitivity for detecting true ADE signals compared to McNemar's test.

Conclusions:

• The PMRM is effective in identifying subpopulation-specific ADE signals from extensive ADE-subpopulation-drug combinations.
• The model controls for both FDR and confounding effects, enhancing the reliability of ADE signal detection.
• This approach improves the ability to detect and prevent ADEs in vulnerable patient populations.