Implementation of OMOP and ConcePTION Common Data Models in CPRD GOLD: Risk of Bleeding and Cardiovascular Outcomes From Anticoagulant Use

The impact of the choice of common data model (CDM) approach on the study results in a real-world evidence (RWE) study is unknown. We aimed to determine potential differences in the results of an RWE study when data were mapped to two different CDMs, ConcePTION and OMOP. With the same instance of CPRD GOLD, data were mapped to both CDMs. Using the same programming steps, we estimated the risk of direct oral anticoagulants (DOACs) vs. vitamin K antagonists (VKAs) on bleeding and cardiovascular (CVD) outcomes in patients with non-valvular atrial fibrillation. Baseline characteristics, incidence rates, and Cox proportional hazards ratios were compared between the analyses. OMOP and ConcePTION mapped study populations included 80,701 (93,350 person-years) and 76,726 exposed persons (100,135 person-years), respectively. DOACs showed no differential risk of CVD compared to VKAs in ConcePTION mapped data (HR 0.99, 95% CI [0.91; 1.08]), but protective effects in OMOP (HR 0.82, 95% CI [0.74; 0.90]). DOACs had a similar increase in risk of stroke (ConcePTION HR 1.19, 95% CI [1.02; 1.37]; OMOP HR 1.10, 95% CI [0.96; 1.26]). No increased risk of major bleeding was identified (ConcePTION HR 0.97, 95% CI [0.84; 1.13]; OMOP HR 0.90, 95% CI [0.78; 1.04]). OMOP gave lower effect estimates for CVD and equivalent risks of stroke and bleeding associated with DOACs use. This study highlights the challenges in repeating the same analysis across the two data CDMs. Differences potentially stem from the cohort construction, the identification of phenotypes in the different CDMs, and the use of imputed variables defined during mapping processes, such as drug exposure duration.