Evaluating analytic models for individually randomized group treatment trials with complex clustering in nested and...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Individually Randomized Group Treatment (IRGT) trials randomize individuals but use shared agents for treatment delivery.
• This shared agent results in post-randomization correlations among participants, complicating analysis.
• Existing analytic models may not adequately address complex clustering in IRGT designs.

Purpose of the Study:

• To evaluate the performance of statistical models for IRGT trials with complex clustering.
• To identify appropriate analytic models for nested and crossed agent designs in IRGT studies.
• To assess type I error rates under various clustering scenarios.

Main Methods:

• A simulation study was conducted to examine candidate analytic models.
• Scenarios included single membership, multiple membership, and single agent settings.
• Designs considered were both nested and crossed, with a continuous outcome.

Main Results:

• For nested designs, type I error rate inflation occurred when multiple membership was not accounted for or when model weights mismatched data generation.
• Analytic models for crossed designs generally maintained nominal type I error rates.
• Imbalance in participants per agent in crossed designs was a key factor affecting error rates.

Conclusions:

• Appropriate statistical modeling is crucial for accurate analysis of IRGT trials.
• Nested designs require careful consideration of multiple membership to avoid inflated type I errors.
• Crossed designs are more robust, but agent imbalance can still impact results.