Multi-Database Methodologies to Enhance Real-World Evidence Generation

Abstract

In using multiple databases, the challenges faced in pharmacoepidemiology are more complicated compared to those that use only a single database. The aim of this thesis was to pinpoint those challenges associated, and to showcase and compare modern strategies employed in these studies. It guides researchers about appropriate methodologies and strategies for performing multi-database studies. It can inform stakeholders using real-world evidence (RWE), including regulatory agencies, health technology assessment agencies and clinical personnel, to make decisions about medicines. Chapter 2 focuses on the transparent conduct of, and the potential strategies involved with multi-database RWE research. First, in Chapter 2.1, a systematic review of how missing data is reported and subsequently handled in such multi-database pharmacoepidemiologic studies, is presented. Afterwards, multiple comparisons of strategies for conducting multi-database RWE studies are investigated. Chapter 2.2 compares the syntactic harmonisation of two electronic healthcare databases provided by a common data model (CDM) to a common protocol-only approach when executing a pharmacoepidemiologic RWE cohort study. This case study investigated the risk of bleeding and cardiovascular disease outcomes in direct oral anticoagulants versus vitamin K antagonists in patients with non-valvular atrial fibrillation. Consistency in study population characteristics, outcome counts, and drug effect estimates was assessed. Chapter 2.3 additionally compares this syntactic harmonisation provided by the ConcePTION CDM to the semantic harmonisation provided by the OMOP CDM in one electronic health record database, for the same parameters. Chapter 3 contains two multi-database RWE studies, where heterogeneity was analyzed. In Chapter 3.1, a drug utilisation study of rivaroxaban used for the management of atherosclerotic cardiovascular disease is presented, using a common protocol approach in two European electronic healthcare databases. In Chapter 3.2, a study investigating the risk of bleeding and stroke after the introduction of direct oral anticoagulants, using data from four European countries is presented. We analysed the persistence of study result heterogeneity after methodological harmonisation as to whether it reveals potential disparities in drug use, clinical practice, or the underlying population across the included databases. Finally, a general discussion of the findings, and recommendations for performing multi-database RWE studies, are made in Chapter 4.