Early Detection of Adverse Drug Reactions in Postmarket Monitoring

Most drugs are associated with some form of adverse drug reactions (ADRs). Understanding the connection between drugs and ADRs is crucial for minimizing patient harm and reducing public healthcare costs. Consequently, there has been sustained interest in correlation analysis within pharmacovigilance and drug development. In the postmarketing phase, the estimated correlation between drugs and their ADRs is affected by both the correlation degree and variability. Therefore, accounting for variability is particularly important when measuring correlations, particularly in the early stage with fewer data points, where variability is typically higher. In this study, we introduce a framework called error-controlled correlation (ECC), which provides correlation estimates while dynamically adjusting for variability. ECC offers a versatile framework that is applicable to any correlation measure. Using the five most widely used correlation measures, we demonstrate ECC’s efficacy in identifying highly correlated drug-ADR pairs while maintaining a controlled type 1 error rate. Experimental results on both real-world and simulated data show that ECC consistently outperforms benchmark methods. Notably, it achieves comparable performance to existing methods with only 1/10th of the data, enabling significantly earlier ADR detection.

History: Accepted by J. Paul Brooks, Area Editor for Applications in Biology, Medicine, & Healthcare.

Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information (https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2024.0585) as well as from the IJOC GitHub software repository (https://github.com/INFORMSJoC/2024.0585). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/.