Background: Warfarin is commonly used in long-term anticoagulation and has a narrow therapeutic index. Dosing is dependent on intrinsic patient variables including cytochrome P450 2C9 (CYP2C9) and VKORC1 polymorphisms. High protein binding and CYP-dependent clearance mechanisms of NSAIDs also affect warfarin serum levels. This study investigates the observed pharmacologic interaction of concurrent NSAID and warfarin use by examining indicators including blood transfusions received and time to therapeutic INR.

Methods: A retrospective, matched cohort study was performed on inpatients initiated on warfarin as part of a heparin “bridge” from January 2006 through January 2016 at an academic medical center. After patients exposed to NSAIDs while on warfarin (“cases”) were identified via the electronic medical record, patients receiving warfarin who were not exposed to NSAIDs (“controls”) were matched randomly to cases by age and gender. Paired t-tests were conducted on cases and controls for outcomes including blood transfusions received during hospitalization, time to therapeutic INR (>2), cumulative warfarin dose administered to achieve therapeutic INR, and the value of the first therapeutic INR. Outcomes between cases who received IV NSAIDs (ketorolac) were also compared against cases who received PO NSAIDs (ibuprofen and naproxen). Multivariate regression models were generated after all univariate analyses to control for race, ethnicity, BMI, renal function, hepatic function, concurrent aspirin use, and other variables.

Results: One hundred and seventy pairs of matched cases and controls were identified. For the following results, corresponding multivariate regression p-values are shown following univariate p-values. Cases who were exposed to NSAIDs received increased blood transfusions during hospitalization compared to controls (p<0.0001; multivariate: p=0.006, OR=4.2). Furthermore, cases who received IV NSAIDs received increased blood transfusions compared to cases who received PO NSAIDs (p<0.0001; multivariate: p=0.002, OR: 7.5). Cases also experienced decreased time to therapeutic INR (3.8 days vs. 4.6 days, p<0.0001; multivariate: p=0.02), decreased cumulative warfarin dose (27 mg vs. 35 mg, p<0.0001; multivariate: p=0.003), and higher first therapeutic INR (2.5 vs. 2.3, p=0.02; multivariate: p=0.04) compared to controls.

Conclusions: Patients exposed to NSAIDs while on warfarin received increased blood transfusions relative to patients who did not receive NSAIDs. These conclusions are further preserved when patients who received IV NSAIDs were compared to those who received PO NSAIDs, signifying a possible potency-mediated effect. These results point to the increased bleeding risk associated with concurrent warfarin and NSAID administration. Notably, cases had significantly decreased time to therapeutic INR, decreased cumulative warfarin dose, and higher first therapeutic INR, suggesting a phenotypic manifestation of NSAIDs delaying warfarin clearance. These findings illustrate a major clinically-relevant adverse drug interaction among two routine medications in the inpatient setting and underscore the importance of careful warfarin dosing in patients already receiving NSAIDs.