Pharmacoepidemiological assessment of drug interactions with vitamin K antagonists

PURPOSE: We present a database of prescription drugs and international normalized ratio (INR) data and the applied methodology for its use to assess drug-drug interactions with vitamin K antagonists (VKAs). We use the putative interaction between VKAs and tramadol as a case study. METHODS: We used a self-controlled case series to estimate the incidence rate ratio (IRR) comparing the rate of INR measurements of ≥4.0 in concomitant tramadol and VKA-exposed periods to VKA-only-exposed periods. Secondary analyses considered specific subgroups, alternative exposure criteria, alternative outcome definitions, and other drugs. RESULTS: We identified 513 VKA users with at least 1 INR measurement ≥4.0 and concomitant tramadol and VKA exposure during the observation period. The overall IRR was 1.80 (95% confidence interval [CI], 1.53-2.10), with a stronger association among users of phenprocoumon compared to warfarin (IRR, 3.37; 95%CI, 2.50-4.53 and IRR, 1.46; 95%CI, 1.20-1.76, respectively). We observed larger IRRs with stricter outcome definitions. Concomitant tramadol and VKA exposure was also associated with an increased rate of low INR measurements (i.e., <1.5; IRR, 1.70; 95%CI, 1.37-2.13). Morphine and, to some extent, oxycodone, penicillin, beta-blockers, and inhaled beta-agonists were associated with high INR. CONCLUSIONS: The approach successfully identified an interaction between tramadol and VKA. However, associations observed for other drugs with no known VKA interaction suggest that the current approach may have too low specificity to be useful as a screening tool, at least for drugs for which time-varying confounding may be present.