Study Designs for Extending Causal Inferences From a Randomized Trial to a Target Population.

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Randomized trials provide robust causal evidence but are often limited in generalizability.
• Extending these findings to broader target populations is crucial for public health and policy.
• Existing methods for causal inference generalization face challenges with diverse study designs.

Purpose of the Study:

• To examine study designs for extending causal inferences from randomized trials to target populations.
• To compare nested and non-nested trial designs for generalizability.
• To investigate the role of sampling probabilities in identifying counterfactual quantities.

Main Methods:

• Analysis of nested trial designs where randomized individuals are within the target population sample.
• Evaluation of non-nested designs, including composite datasets combining trial and external non-randomized data.
• Application of the g-formula and inverse probability weighting for identifying counterfactual outcome means.

Main Results:

• The identifiability of counterfactual quantities is contingent upon knowledge of non-randomized sampling probabilities.
• Both nested and non-nested designs have specific conditions for valid causal inference generalization.
• The probability of trial participation can be identified and estimated based on sampling properties.

Conclusions:

• Study design critically influences the ability to generalize causal inferences from randomized trials.
• Understanding sampling mechanisms is key for robust causal transportability.
• The g-formula and inverse probability weighting are valuable tools for causal inference in complex designs.