133 - Remote, bivariate expert elicitation to determine the prior probability distribution for sample size calculation in a Bayesian non-inferiority Croup Dosing Trial

Publication Number: 133.6127

Alex O. Aregbesola, University of Manitoba, Winnipeg, MB, Canada; Apoorva Gangwani, The Children's Hospital Research Institute of Manitoba, Winnipeg, MB, Canada; Arlene Jiang, The Hospital for Sick Children, Toronto, ON, Canada; Terry P. Klassen, University of Saskatchewan, Saskatoon, SK, Canada; Anna Heath, The Hospital for Sick Children, Toronto, ON, Canada; Amy Plint, University of Ottawa Faculty of Medicine, Ottawa, ON, Canada; Elisabete Doyle, Dr. Elisabete Doyle Medical Corporation, Winnipeg, MB, Canada; William Craig, University of Alberta Faculty of Medicine and Dentistry, Edmonton, AB, Canada; Mohamed Eltorki, university of calgary, Calgary, AB, Canada; Yongdong Ouyang, The Hospital for Sick Children, Toronto, ON, Canada; Banke Oketola, Max Rady College of Medicine, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB, Canada; Hoda Badran, CHRIM, Winnipeg, MB, Canada

Background: In Bayesian clinical trials, prior distributions must be specified for key parameters. When existing evidence is limited, expert elicitation can be used to construct these distributions. Traditional elicitation involves face-to-face interaction and intensive training, which is costly and time-consuming. This study conducted remote expert elicitation to generate a bivariate prior distribution for the Croup Dosing Trial, assessing the non-inferiority of 0.15 mg/kg dexamethasone compared to the standard 0.60 mg/kg dose in reducing return visits to the emergency department (ED) within 7 days for children with croup.

Objective: The study aimed to elicit a prior distribution to calculate the sample size for a Bayesian non-inferiority trial, evaluating the effectiveness of a lower dose of dexamethasone (0.15mg/kg instead of 0.60 mg/kg) in reducing return emergency department (ED) visits within 7 days.

Design/Methods: We conducted three remote workshops to elicit expert beliefs on the efficacy of the two doses of dexamethasone. Each workshop consisted of two survey rounds, separated by a group discussion. Prior to the workshop, experts reviewed the same current literature that was provided on the effects of the two doses of dexamethasone. Beliefs were aggregated using expert-specific bivariate distributions with latent effects. The aggregated distribution, along with the surveyed non-inferiority margin, determined the sample size for the Bayesian non-inferiority trial design.

Results: The elicitation generated a prior distribution centered at 6% for the 0.60 mg/kg dose and 8% for the 0.15 mg/kg dose. The aggregated prior distribution produced a sample size of 1700, based on a non-inferiority margin of 4.88%.

Conclusion(s): We elicited a prior distribution that incorporated past evidence and expert opinion. The elicited prior is consistent with previous literature on the efficacy of the dexamethasone doses in treating croup. Our approach demonstrates the feasibility of remotely eliciting bivariate distributions to design clinical trials.

Table 1: Baseline characteristics of the expert elicitation participants.
Table 1 - Baseline characteristics.jpegTable 1 presents the professional experience characteristics of experts from our survey. These experts reported between 11 and 38 years of medical practice. In the past year, all physicians had prescribed dexamethasone at a dose of 0.60 mg/kg, while only two physicians had prescribed a dose of 0.15 mg/kg.

Figure 1: Deidentified boxplots of expert opinion after Round 1
Figure 1 shows a deidentified boxplot generated after the first round to facilitate group discussion. These discussions allowed experts to comment on their reasoning behind the estimates they chose. Between survey rounds, experts revisited their interpretation of the lower and upper plausible values and considered how the primary outcome relates to the clinical progression of croup.

Figure 2: Aggregated Prior Distribution of Probability for Emergency Department (ED) Visits
Figure 2 displays the aggregated prior distributions from all experts across each survey round. These graphs illustrate a reduction in variability in responses from round 1 to round 2. The aggregated distributions from round 2 were used as the final prior distributions for determining sample size.

Table 1: Baseline characteristics of the expert elicitation participants.
Table 1 - Baseline characteristics.jpegTable 1 presents the professional experience characteristics of experts from our survey. These experts reported between 11 and 38 years of medical practice. In the past year, all physicians had prescribed dexamethasone at a dose of 0.60 mg/kg, while only two physicians had prescribed a dose of 0.15 mg/kg.

Figure 1: Deidentified boxplots of expert opinion after Round 1
Figure 1 shows a deidentified boxplot generated after the first round to facilitate group discussion. These discussions allowed experts to comment on their reasoning behind the estimates they chose. Between survey rounds, experts revisited their interpretation of the lower and upper plausible values and considered how the primary outcome relates to the clinical progression of croup.

Figure 2: Aggregated Prior Distribution of Probability for Emergency Department (ED) Visits
Figure 2 displays the aggregated prior distributions from all experts across each survey round. These graphs illustrate a reduction in variability in responses from round 1 to round 2. The aggregated distributions from round 2 were used as the final prior distributions for determining sample size.