Comparative treatment effects of direct oral anticoagulants versus warfarin for nonvalvular atrial fibrillation in clinical practice vary over time due to channelling

AIM: Temporal shifts in patient characteristics can influence comparative treatment effects within a certain indication. We investigated patient characteristics and comparative effectiveness and safety for direct oral anticoagulants (DOACs) and warfarin over the time period where DOACs replaced warfarin as preferred stroke prophylaxis in patients with nonvalvular atrial fibrillation (NVAF). METHODS: We applied the cohort design to the Clinical Practice Research Datalink Aurum and created 12 sequential subcohorts of patients with NVAF between 2011 and 2023. Patients with a DOAC or warfarin prescription following a 1-year prescription-free period were included. Baseline characteristics were compared between subcohorts. Survival analysis was used to evaluate the risk of ischaemic stroke (effectiveness outcome) and major bleeds (safety outcome). RESULTS: We included 207 393 patients with NVAF (45% women, 77% DOAC users, mean age 76 years). In early subcohorts, a history of stroke/transient ischaemic attack was more common among DOAC users. In later subcohorts, diabetes, deep venous thrombosis/pulmonary embolism and antihypertensive use were more common among warfarin users. History of other cardiovascular disease, renal failure and diuretic use were consistently higher among warfarin users. Over time, the crude stroke hazard ratio for DOAC users compared with warfarin users shifted from 2.66 (95% CI 2.00-3.52) to 0.62 (0.23-1.67). This trend was attenuated in adjusted analyses. No such trends were seen for major bleeds. CONCLUSION: DOACs had unfavourable effectiveness estimates in earlier subcohorts, potentially due to preferential prescribing to higher risk patients. Early observational studies may encounter initial biases, underscoring the need for robust methodological approaches when evaluating newly introduced medicines.