Gastrointestinal bleeding risk among antithrombotics users: a propensity score matched study of drug classes

Background: Antithrombotic drug-drug interactions (DDIs) impact patient safety. This study used causal inference framework to examine the association between interacting drugs, direct oral anticoagulants (DOACs), antiplatelets, warfarin and incident gastrointestinal (GI) bleeding risk. Research design and methods: Data from Merative MarketScan commercial and Medicare claims (2013–2021), enriched with Micromedex data were used. Adults (≥18 years) with antithrombotics claims were included. Incident GI bleeding was primary outcome. Adaptive post-double selection using least absolute shrinkage and selection operator (LASSO) was implemented for variable selection. Following propensity score matching, final estimates were obtained from weighted logistic models. Results: The final matched cohorts included 368,934 patients for DOACs vs Antiplatelets; 261,184 for DOACs vs warfarin and 192,955 for antiplatelets vs warfarin. Baseline characteristics were well balanced (standardized mean difference (SMD) <0.05). Compared to antiplatelets, DOACs were associated with a significant reduction in incident GI bleeding risk (odds ratio (OR): 0.94; 95% CI: 0.91–0.98; P_FDR = 0.006). None of the interacting drugs reached statistical significance, bleeding risk was driven by age (>70 years) and comorbidities. Conclusions: Within a causal inference framework, DOACs exposure was associated with less risk of GI bleeding compared to antiplatelets. However, class-level grouping may mask individual drug-specific metabolic risks. IRB: The protocol (STUDY00016941) was reviewed and received exemption determination by the University of Minnesota Institutional Review Board.