On causation in therapeutic research: observational studies, randomised experiments and instrumental variable...

Area of Science:

• Epidemiology
• Biostatistics
• Health Economics

Background:

• Causal inference is crucial for evaluating therapeutic interventions.
• Observational studies often lack group exchangeability, hindering causal claims.
• Randomized controlled trials (RCTs) provide strong causal evidence through expected exchangeability.

Purpose of the Study:

• To explore the utility of instrumental variables (IV) in strengthening causal claims from observational data.
• To address the limitations of observational studies in establishing causality for therapeutic effects.

Main Methods:

• The study discusses the counterfactual theory of causation.
• It examines the role of exchangeability in comparative studies.
• The application and assumptions of instrumental variables (IV) are detailed as a method to mimic RCTs.

Main Results:

• Observational studies face challenges in establishing causal links due to non-exchangeable groups.
• Randomization in RCTs ensures expected exchangeability, supporting straightforward causal interpretation.
• Instrumental variables can mitigate some issues of observational data by finding variables associated with treatment but not directly with outcomes.

Conclusions:

• Instrumental variables provide a valuable, though assumption-heavy, approach to causal inference in observational studies.
• This method offers a more appealing alternative than avoiding causal statements altogether.
• The use of IVs can enhance the reliability of causal claims for therapeutic interventions.