Efficient and robust methods for causally interpretable meta-analysis: Transporting inferences from multiple...

Area of Science:

• Biostatistics
• Epidemiology
• Causal Inference

Background:

• Meta-analyses typically synthesize findings but often lack causal interpretability for new populations.
• Transporting causal inferences across different populations presents significant methodological challenges.

Purpose of the Study:

• To develop methods for causally interpretable meta-analyses.
• To enable the transport of causal inferences from randomized trials to a target population where experimental data may be absent.

Main Methods:

• Investigated identifiability conditions for causal inference.
• Derived implications for observed data laws.
• Developed an estimator for potential outcome means in the target population using trial and target population covariate data.

Main Results:

• Obtained identification results for transporting causal inferences.
• Proposed a doubly robust estimator, consistent and asymptotically normal under weaker model assumptions.
• Demonstrated finite sample properties via simulation and real-world data.

Conclusions:

• The proposed methods facilitate robust causal inference in target populations using data from multiple randomized trials.
• The doubly robust estimator offers a reliable approach for causal transport when direct experimental data is unavailable.