A Bayesian adaptive design for clinical trials in rare diseases

Area of Science:

• Clinical Trials
• Biostatistics
• Rare Diseases

Background:

• Developing treatments for rare diseases is difficult due to small patient populations.
• Traditional fixed randomized designs may not be feasible for rare disease trials.
• An alternative goal is to maximize patient success within the trial itself.

Purpose of the Study:

• To propose a novel randomized response-adaptive trial design.
• To maximize the total number of patient successes in a clinical trial.
• To penalize under-recruitment to treatment arms while ensuring randomization.

Main Methods:

• Utilized finite-horizon Markov decision processes and dynamic programming (DP).
• Developed a novel randomized response-adaptive design for two-armed trials with binary endpoints.
• Evaluated performance measures through extensive simulation studies.

Main Results:

• The proposed design significantly increases patient allocation to the superior treatment arm compared to fixed designs.
• It substantially improves statistical power relative to optimal DP designs.
• The design demonstrates minimal bias and mean squared error in treatment effect estimation.

Conclusions:

• The novel adaptive design effectively maximizes patient successes in rare disease trials.
• It offers practical advantages, including full randomization, addressing barriers to implementing bandit models in clinical practice.
• This approach enhances treatment development for rare diseases by optimizing within-trial outcomes.