An Expected Score Approach to Ordinal Outcomes in a Bayesian, Response Adaptive, Randomized Trial.

Area of Science:

• Neurology
• Clinical Trials
• Biostatistics

Background:

• Ordinal outcomes are prevalent in medical research but pose analytical challenges.
• Benzodiazepine-refractory status epilepticus requires effective treatment strategies.
• Previous studies suggest a bimodal distribution for ordinal outcomes in this context.

Purpose of the Study:

• To propose and evaluate a novel Bayesian analytic approach for ordinal outcomes with non-proportional effects.
• To compare the proposed method against traditional frequentist approaches in simulated scenarios.
• To assess the utility of this approach in the KESETT trial for status epilepticus.

Main Methods:

• A Bayesian adaptive clinical trial design (KESETT study) was simulated.
• An ordinal primary outcome (60 minutes) with a 5-point scale was used.
• The proposed method analyzes expected scores from posterior distributions, avoiding proportionality assumptions.

Main Results:

• The novel Bayesian approach effectively controls Type I error and maintains statistical power.
• This method demonstrates a significant reduction in required sample size compared to non-parametric tests.
• The approach is interpretable and leverages the full ordinal scale.

Conclusions:

• The proposed Bayesian analytic method is suitable for ordinal outcomes with non-proportional effects.
• This approach offers a robust and efficient alternative to traditional statistical methods in clinical trials.
• The method is particularly relevant for complex medical conditions like status epilepticus.