(233) Identification of Unrecognized Associations Between Medications and Gastrointestinal Bleeding by Screening Large Administrative Health Data

Publication Number: 893

Background: Anticoagulants, antiplatelets, and NSAIDs are known to increase the risk of gastrointestinal bleeding (GIB). Additionally, there are other medications, such as selective serotonin-reuptake inhibitors (SSRIs), that have inconsistently demonstrated association with GIB. Furthermore, there may be additional medications that influence GIB, but have not yet been recognized.

Objectives: The aim of this study was to apply novel biostatistical techniques that identify unanticipated medication effects to confirm previously reported medication-GIB associations and to identify medications with unrecognized effects on GIB.

Methods: Using Optum Clinformatics, we performed a pharmacopeia-wide association study (PWAS), which is a variation of a case-control study. Individuals diagnosed with GIB from 2010 – 2021 during a hospitalization were classified as cases. A 5% random sample of cases were matched to four controls on age and sex. We then identified outpatient medication use through pharmacy claims. Individuals were classified as exposed to a given medication if they filled a prescription that provided at least one dose between 30 and 5 days prior to the index date. The association of each medication with GIB was assessed using conditional logistic regression, adjusted for comorbidities, number of concurrent medications, and a high-dimensional propensity score. We then applied medication class enrichment analysis (MCEA), a signal-to-noise enhancement algorithm, to reduce false positive results by identifying groups of medications with similar mechanisms of action and consistent association with GIB.

Results: We identified 8579 cases with GIB and 34,316 matched controls (mean age 72.0, 50.6% female). Of 283 tested medications, warfarin, rivaroxaban, apixaban, clopidogrel, and naproxen were among the most strongly associated medications with GIB (ORs 1.90, 2.78, 1.73, 1.44, and 2.21, respectively). Omeprazole demonstrated a strong inverse association (OR 0.73), and none of the SSRIs were associated with GIB. Previously unrecognized associations included a strong association between metronidazole and GIB (OR 4.89) and strong inverse associations with GIB for empagliflozin (OR 0.20) and anastrozole (OR 0.26). Each described association was below the Bonferroni threshold. MCEA demonstrated consistent within-pharmacologic class associations for rivaroxaban, apixaban, clopidogrel, and empagliflozin.

Conclusions: Using novel biostatistical approaches, we confirmed known medication-GIB associations with anticoagulants, antiplatelets, and NSAIDs. We did not identify an association between SSRIs and GIB. Finally, we demonstrated previously unrecognized associations with metronidazole, empagliflozin, and anastrozole.