Using Bayesian adaptive designs to improve phase III trials: a respiratory care example.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Critical Care Medicine

Background:

• Traditional clinical trial designs can be inefficient, requiring large patient populations and extended timelines.
• Bayesian adaptive designs offer a flexible alternative to improve trial efficiency and expedite evidence generation.
• Phase III clinical trials in critical care present unique challenges that may benefit from adaptive methodologies.

Purpose of the Study:

• To construct and evaluate Bayesian adaptive designs for phase III critical care trials.
• To assess the impact of Bayesian adaptive designs on trial efficiency, including sample size and duration.
• To demonstrate the practical application of Bayesian adaptive designs using real clinical trial data.

Main Methods:

• Re-designed the High Frequency OSCillation in Acute Respiratory distress syndrome (OSCAR) trial using Bayesian adaptive methods.
• Constructed and simulated operating characteristics of five alternative Bayesian adaptive designs.
• Performed virtual re-executions of the OSCAR trial using the developed Bayesian designs.

Main Results:

• Identified a preferred Bayesian adaptive design with similar power to the original OSCAR trial but requiring fewer patients on average.
• Virtual re-executions indicated that Bayesian sequential designs could lead to reduced sample sizes and earlier trial completion.
• Bayesian sequential approaches yielded similar trial conclusions to the original OSCAR trial design.

Conclusions:

• Bayesian adaptive designs are feasible for phase III critical care trials, as demonstrated by the OSCAR trial case study.
• Implementing Bayesian adaptive designs in the OSCAR trial could have resulted in a smaller sample size and earlier termination with comparable conclusions.
• Wider adoption of Bayesian adaptive approaches is recommended for phase III clinical trials to enhance efficiency and reduce patient burden.