Until the 1940s, the development of new treatments relied on NRS. After that time, there was increasing recognition that anecdotal reports based on clinical practice observations were often misleading. This led to a near-total replacement of the prior nonrandomized approach with the use of randomized, controlled clinical trials (RCT). Indeed, the history of medicine is rife with examples whereby observational data have been misleading even with established clinical practices, and which are only uncovered after the same hypothesis is tested in an RCT. This reinforces the widely held notion about NRS that no matter how large in scale or sophisticated in analysis, the risk of bias (including misspecification, selection, reporting, analysis, and confounding, among others) will limit certainty in causal inference. Conversely, proponents of NRS RWE advocate that mechanistic trials may often not be fully representative of real-life situations because they employ strict, protocol-defined inclusion criteria to identify eligible patients – that is to say, directness in the applicability of the studied intervention effects to the applied population. This could mean that some patients with the condition of interest may be excluded based on characteristics such as disease severity, age, comorbidities, or the use of concomitant medications. Since a few years, severe asthma is a great example of both bias in selection of population enrolled in registration trials for biologics and how the registries provided useful information about both diagnosis and management. Finally, they revealed how to better define more severe phenotypes and how oral corticosteroids (OCS) impact in severe asthma treatment in real life.