Stroke prevention in atrial fibrillation: re-defining 'real-world data' within the broader data universe.

Real-world data (RWD) has been defined as data generated outside of traditional randomized clinical trials (RCTs). Though RWD has received increasing attention from regulatory authorities and professional societies, dividing evidence into that derived from 'real-world' vs. 'non-real-world' sources provides only one element of a much larger framework for evidence evaluation. Evidence should be evaluated on the source of the data, the method of treatment allocation (whether any intervention being evaluated was assigned or simply observed as used in practice) and the context in which the evidence was generated (overall study design). Under this framework, RWD refers only to data source, and a study incorporates RWD when it primarily uses data collected for non-research purposes, such as insurance claims data or the electronic health record, regardless of study design. Separation of study design, data source, and context enables parallel evaluation of two critical elements: (i) whether a study can support claims of causal inference, which can be assured with a high degree of confidence only in studies where patients are assigned treatments by protocol; and (ii) whether the study population and clinical context mirror clinical practice, a strength of observational studies using data from clinical practice or administrative claims. In this review, we describe the strengths and weaknesses of observational and non-observational studies, and studies involving RWD and non-RWD, through the lens of anticoagulation for atrial fibrillation (AF). Observational studies employing RWD are useful for describing how oral anticoagulants are used in clinical practice, but generally cannot be used to make claims regarding comparative treatment effects. Questions regarding treatment effect generally are best answered through an RCT, and additional pragmatic RCTs are needed to compare different antithrombotic agents for the prevention of thrombotic events in AF.