Avoidable Bias in Real-World Evidence Studies on Treatment Effects: How to Detect and Prevent It

Background: Real-world evidence (RWE) studies of treatment effects are subject to multiple sources of bias. While confounding and data quality are often featured as major challenges in RWE generation, it has been shown that many RWE studies yield biased results because of self-inflicted and avoidable errors in design, leading to inclusion or exclusion of immortal person-time, or depletion of outcome-susceptible individuals, among others. The complexity of data, study design, and analytical methods makes detecting and preventing such bias challenging.

Objectives: The main objectives of this symposium are (1) to provide an overview of major sources of avoidable bias in RWE drug safety or effectiveness studies, (2) to present strategies that will help the audience identify such bias in RWE studies, and (3) to present and discuss current approaches to avoid self-inflicted bias when designing RWE studies. This symposium will be valuable to any ISPE member who needs to either critically evaluate or design RWE studies on drug safety or effectiveness.

Description: Dr. Bykov will introduce the concept of avoidable bias in RWE drug safety or effectiveness studies (7 mins). Dr. Patorno will discuss four major sources of avoidable bias: bias due to immortal person-time, depletion of outcome-susceptible individuals, reverse causation, and differential surveillance, and how to identify these biases in published RWE (18 mins). Subsequently, the presenters will focus on current strategies to avoid self-inflicted bias in RWE studies. First, Dr. Fu will present the advantages of target trial emulation, including clone-sensor-weighting, when trying to avoid self-inflicted bias (15 mins). Then Dr. Franklin will discuss advantages and limitations of an active-comparator, new-user design (15 mins). She will be followed by Dr. Suissa who will present his work and strategies to avoid bias due to immortal person-time, including prevalent new user design (15 mins). At the end, Dr. Bykov will provide final remarks (5 mins) and the panelists will take questions and discussion points from the audience (15 mins). In addition to providing a forum for discussing different approaches to study design, this symposium will provide a primer on self-inflicted bias in RWE studies, and a toolbox of approaches that could help avoid it.