Model driven method for exploring individual and confounding effects in spontaneous adverse event reporting databases

Abstract

ABSTRACT Background Spontaneous Adverse Event Reporting (SAER) databases play a crucial role in post-marketing drug surveillance. However, the traditional model-free disproportionality analysis has been challenged by the insufficiency in investigating subgroup and confounders. These issues result in significant low-precision and biases in data mining for SAER. Methods The Model-Driven Reporting Odds Ratio (MD-ROR) was proposed to bridge the gap between SAER database and explainable models for exploring individual and confounding effects. MD-ROR is grounded in a well-designed model, rather than a 2 × 2 cross table, for estimating AE-drug signals. Consequently, individual and confounding effects can be parameterized based on these models. We employed simulation data and the FDA Adverse Event Reporting System (FAERS) database. Result The simulated data indicated the subgroup effects estimated by MD-ROR were unbiased and efficient. Moreover, the adjusted-MD-ROR demonstrated greater robustness against confounding biases than the crude ROR. Applying our method to the FAERS database suggested higher occurrences of drug interactions and cardiac adverse events induced by Midazolam in females compared to males. Conclusion The study underscored that MD-ROR holds promise as a method for investigating individual and confounding effects in SAER databases.