A Tree-Based Scan Statistic for Detecting Signals of Drug-Drug Interactions in Spontaneous Reporting Databases

Abstract

The concomitant use of multiple drugs increases the risk of adverse events (AEs) due to drug-drug interactions (DDIs), which remain challenging to identify since clinical trials primarily focus on individual drugs, necessitating postmarket safety monitoring through spontaneous reporting systems. Although several statistical methodologies have been proposed to detect DDI signals with disproportionately high reporting rates, existing methods inadequately account for the hierarchical structure of AEs and potential reporting bias. To address these limitations, we developed a statistical methodology that incorporates the hierarchical structure of AEs using tree-based scan statistics while mitigating reporting bias by assuming DDIs follow a multiplicative interaction model. In simulation studies, our proposed method effectively controlled type I error rate at prespecified significance levels across all simulation scenarios and demonstrated consistent performance in power, sensitivity, and false discovery rate, even with reporting bias present. This novel tree-based scan statistic methodology for detecting DDI signals that accounts for both hierarchical AE structure and potential reporting bias can serve as a valuable tool for postmarket drug safety surveillance.