Objectives: To review advances in patient-relevant causal effects.
Description: The intention-to-treat (ITT) analysis for randomized controlled trials provides an unbiased estimate of the causal effect of treatment assignment. This is relevant for policy makers because patient-adherence is essential in order to know the programme’s overall effectiveness. However, an individual deciding whether to take treatment or not, needs to know if the treatment is effective if taken (patient-relevant analysis), not the causal effect averaged over those who do and do not take treatment.
Although 'per protocol' and 'as treated' analyses are often used, the underlying assumptions are almost always violated. Newer methods based on instrumental variables and principal stratification are gaining popularity, and can be helpful. Cochrane has been a strong proponent of patient-relevant outcomes, and patient-relevant analyses would appear to be the next logical step.
The purpose of this workshop is to explain the strengths, weaknesses and underlying assumptions when estimating patient-relevant causal effects. Using data from two published studies, we will review the different questions being addressed by different methods, how to define 'adherence', and underlying assumptions of each method.