Generalizability Analyses with a Partially Nested Trial Design: The Necrotizing Enterocolitis Surgery Trial.

Area of Science:

• Clinical Trials
• Biostatistics
• Epidemiology

Background:

• Partially nested trial designs present unique challenges for generalizability analyses.
• These designs occur when data collection varies for randomized and non-randomized individuals or across trial phases.
• The Necrotizing Enterocolitis Surgery Trial exemplifies this design, with data collection changes between study phases.

Purpose of the Study:

• To propose and evaluate methods for generalizability analyses in partially nested trial designs.
• To enable causal inference in the target population of all trial-eligible individuals, including both randomized and non-randomized participants.
• To address the complexities arising from incomplete data on non-randomized individuals.

Main Methods:

• Develop methods for generalizability analyses tailored to partially nested trial designs.
• Define identification conditions for causal estimands.
• Propose estimators that utilize data from both nested and non-nested parts of the trial.
• Evaluate proposed estimators through a simulation study.

Main Results:

• The proposed methods provide valid causal estimates for the target population.
• Simulation results demonstrate the performance of the proposed estimators.
• The methods are applicable even when data on non-randomized individuals are limited in duration or scope.

Conclusions:

• Generalizability analyses can be effectively conducted in partially nested trial designs.
• The proposed methods enhance causal inference by integrating data from all trial components.
• These techniques are crucial for maximizing the utility of complex clinical trial data.