Observational studies

It is widely accepted that the randomised controlled trial has the potential to provide the best evidence for the effectiveness of a new treatment. There can be some special circumstances however where conducting a randomised trial is problematic and where an external control group can be used as a basis to evaluate efficacy and safety. Such non‐randomised studies however are susceptible to a range of different biases and this chapter explores the nature of those biases and how to address them from a statistical point of view. Selection bias results in imbalances in known and unknown baseline factors.