Remote, bivariate prior elicitation for a Bayesian non-inferiority randomized controlled trial

Area of Science:

• Bayesian statistics
• Clinical trial design
• Pediatric emergency medicine

Background:

• Bayesian clinical trials require prior distributions, often derived from expert elicitation.
• Remote elicitation offers a feasible alternative to traditional in-person methods, especially when evidence is limited.
• This study focused on eliciting priors for a trial comparing two dexamethasone doses for pediatric croup.

Purpose of the Study:

• To remotely elicit a bivariate prior distribution for the Croup Dosing Trial.
• To assess the efficacy of two dexamethasone doses (0.15 mg/kg vs. 0.60 mg/kg) in reducing emergency department return visits for croup.
• To determine the sample size for a Bayesian non-inferiority trial design.

Main Methods:

• Conducted three remote workshops with emergency medicine physicians.
• Utilized a two-round survey with group discussion, preceded by literature review.
• Aggregated expert beliefs using bivariate distributions with latent effects to inform trial design.

Main Results:

• Twelve physicians participated in the remote elicitation.
• Generated a prior distribution with estimated efficacy rates of 6% for the 0.60 mg/kg dose and 8% for the 0.15 mg/kg dose.
• The aggregated prior distribution determined a required sample size of 1850 participants for a 4% non-inferiority margin.

Conclusions:

• Successfully elicited a prior distribution integrating existing evidence and expert judgment.
• The elicited prior aligns with existing literature on dexamethasone efficacy in croup treatment.
• Demonstrated the feasibility and utility of remote bivariate distribution elicitation for clinical trial design.