Methodological Challenges in the Identification of Drug-Drug Interactions Using Spontaneous Reporting System

Abstract

Introduction: It has been shown that reporting bias can distort estimates of disproportionate reporting in the single drug setting, however, their influence is unclear when screening spontaneous reporting databases for drug-drug interactions (DDI). Antiviral medications were repurposed to treat COVID-19 during the pandemic period which may have introduced reporting bias. Its impact on DDI signal detection, particularly when using restricted comparator designs for systematic bias mitigation, remains unclear. To investigate this potential we conducted a retrospective bias analysis on a well known DDI.

Methods: We used data from the United States Food and Drug Administration Adverse Event Reporting System (2000Q3-2023Q3) to evaluate changes in disproportionality estimates for lopinavir/ritonavir and atorvastatin with myopathy/rhabdomyolysis. We computed signals using three methods: Concomitant Signal Score (CSS), Omega Shrinkage, and the extended-Bayesian Confidence Propagation Neural Network (BCPNN). Comparisons were conducted across unrestricted and active comparator reference sets (all statins and CYP3A4 statins), pre- and during-pandemic, change in estimates (ACiE) was calculated to quantify differences.

Results: In the unrestricted comparator design, Omega and BCPNN estimates decreased during-pandemic when including Paxlovid (ACiE: -0.37 and -0.24, respectively), potentially by increased background reporting, but increased when excluding Paxlovid (ACiE: 0.84 and 0.16, respectively). In contrast, signal strength increased in all active comparator analyses, particularly with CYP3A4 statins (Omega ACiE: 1.05; BCPNN ACiE: 0.31). CSS results showed a similar trend.

Conclusion: Changes in antiviral indications in response to the COVID-19 pandemic may have altered reporting patterns affecting DDI signal detection.