Preference-Informed Cluster Randomized Design for Pragmatic Clinical Trials

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Public Health Research

Background:

• Cluster randomized trials (CRTs) are essential for pragmatic research but vulnerable to treatment non-adherence.
• Cluster-level preferences can lead to deviations from assigned treatments, impacting trial validity.
• Kawasaki Disease trials face challenges with institutional preferences influencing participation and adherence.

Purpose of the Study:

• To propose and evaluate a Bayesian hierarchical model for CRTs with non-adherence.
• To address treatment switching influenced by cluster-level preferences in pragmatic trials.
• To improve the analysis of CRTs when adherence to randomization is unrealistic.

Main Methods:

• Developed a Preference-Informed Cluster Randomized Design (PICRD) using a Bayesian hierarchical model.
• Explicitly incorporated cluster-level treatment switching into the analysis framework.
• Conducted simulation studies to assess model performance under various switching proportions and effect sizes.

Main Results:

• The PICRD model demonstrated superior performance compared to per-protocol analyses.
• PICRD maintained higher statistical power for detecting treatment effects.
• The model produced narrower credible intervals and more stable bias and RMSE metrics.

Conclusions:

• The PICRD approach offers a flexible and robust solution for analyzing CRTs in pragmatic settings.
• Explicitly modeling preference and treatment switching enhances the validity of CRT findings.
• This method is crucial for real-world trials where perfect adherence is uncommon.