In this methodological study, we will focus on patients with a heart condition called non-valvular atrial fibrillation who are taking blood-thinning drugs and calculate their risk of several acute events for those using direct oral anticoagulants (a novel type of blood thinning drug) vs. those using longer-existing Vitamin K antagonists. Although this has been studied before, including in CPRD, a new element is being added. Assessing whether drugs work in daily clinical practice and are safe is important from a public health perspective. To increase the number of people included in such safety studies, thereby increasing the power, nowadays often different data sources are used. This helps to look at specific groups of patients and find any rare side effects. However, there is a challenge as different data sources store medical data in different data formats. This can cause problems when we want to compare or combine data from different sources. There are several methods to deal with such differences, including methods that alter the original data into a standard format. Therefore, we will also compare the results we find when the study is performed under three different scenarios for handling data structure differences: one that sticks to a common study protocol (no changes made to the original data), one which adjust the original data to fit a common data format fine-tuned to the study protocol and one that changes the original data to a common data format independent of the study protocol.
In recent years many studies have addressed the question whether the risk of major bleeding events among users of Direct Oral Anticoagulants (DOACs) is different compared to use of traditional Vitamin K Antagonists (VKAs). In this project, we will focus on patients with a diagnosis of non-valvular atrial fibrillation and compare the risk of stroke, major bleeding, cardiovascular disease, glaucoma and hip fractures (the latter two are negative outcomes) after the use of DOACs or VKAs. Cox proportional hazards regression models will be used to calculate hazard ratios (HRs). An additional,methodological objective is to study the impact of using different data harmonisation methods. Nowadays, use of multiple health databases is a preferred method for generating evidence on the safety and effectiveness of licenced medicines, enabling researchers to compare drug safety and effectiveness between countries, regions and healthcare systems; investigate specific groups of patients or identify rarer outcomes. For these reasons, they are deemed essential by national and international regulators to assess the clinical effects of drugs. However, these larger studies have an additional number of complexities. In part, these are due heterogeneity amongst data sources, a challenge only amplified when using multiple internationally distributed databases. Harmonisation methods such as the use of a common protocol (CP) and/or common data model (CDM) can mitigate bias and improve precision. A CDM can broadly be characterised as either developed in accordance with a protocol to fit study-specific data (i.e. a study specific CDM) or prior to a specific research protocol (i.e. a general CDM). We will compare the estimates from cox proportional hazards regression models and descriptive statistics (incident rates) of the case study after the application of a study-specific CDM with syntactic harmonisation, a general-use CDM and the use of a common-protocol only.
Risk of stroke (haemorrhagic, ischaemic, or unclassified but not transient ischaemic attack), major bleeding, cardiovascular disease (acute myocardial infarction, acute coronary syndrome, any stroke, other ischemic heart disease), glaucoma (as a negative outcome) and hip fractures (as a negative outcome).
In the exposure
Olaf Klungel - Chief Investigator - Utrecht University
Patrick Souverein - Corresponding Applicant - Utrecht University
Daniel Prieto-Alhambra - Collaborator - University of Oxford
Helga Gardarsdottir - Collaborator - Utrecht University
Marloes Bazelier - Collaborator - Utrecht University
Miriam Sturkenboom - Collaborator - University Medical Centre Utrecht
Nicholas Hunt - Collaborator - Utrecht University