PURPOSE: To provide an introduction and applied examples of how probabilistic quantitative bias analysis (QBA) can be used to measure uncertainty in treatment effect estimates derived from real-world data (RWD) in the presence of unmeasured confounding, measurement error and selection bias.

DESCRIPTION: Analyses of RWD are increasingly being used to support HTA and regulatory submissions. Last year the FDA provided guidance for its use in submissions to the agency. The effective use of RWD to support regulatory or HTA submissions relies on being able to generate unbiased estimates of treatment effects. However, decision makers are often concerned with possible sources of bias in evidence generated from RWD including selection bias, measurement error and confounding. Statistical adjustments abate some of these concerns but the completeness of RWD with respect to variables that would allow for the measurement of bias remains a concern. QBA is a method that allows for the quantitative measurement of selection bias, unmeasured confounding, and measurement error on the direction and magnitude of an estimated treatment effect. The workshop will first provide an overview of the methods underlying the approach (10 minutes). We will then present an application of QBA by describing analyses of the Flatiron Health Analytic Database for treatment of non-small cell lung cancer (20 minutes). Finally, other worked examples will be presented based on common criticism encountered in FDA reviews of evidence from RWD (15 minutes). The workshop will also review approaches for combining QBA with methods to handle missing data, two problems commonly encountered together in RWD. Statisticians, and regulatory or HTA reviewers will benefits from attending this workshop gaining exposure to a technique that will become more commonly applied as access and use of RWD in submissions increases.