Bayesian adaptive decision-theoretic designs for multi-arm multi-stage clinical trials.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Decision Theory

Background:

• Multi-arm multi-stage (MAMS) trials offer advantages over single- or two-arm trials for investigating multiple drugs simultaneously.
• Adaptive designs allow for modifications during a trial based on accumulating data, potentially increasing efficiency.

Purpose of the Study:

• To propose a generic Bayesian adaptive decision-theoretic design for multi-arm multi-stage clinical trials with K arms.
• To define a loss function that balances patient accrual costs and costs of incorrect trial decisions.
• To evaluate the design's performance in terms of decision-making accuracy and efficiency.

Main Methods:

• A Bayesian adaptive framework is employed, making decisions at each stage based on expected loss reduction.
• A novel loss function is defined, incorporating accrual costs and decision-related costs, which is computationally efficient.
• Frequentist operating characteristics are evaluated for binary outcomes, with and without a control arm.

Main Results:

• The proposed Bayesian adaptive design increases the probability of making a correct decision at the trial's conclusion.
• The design demonstrates improved performance compared to nonadaptive and adaptive two-stage designs.
• The loss function estimation is computationally manageable, even for trials with more than two arms (K>2).

Conclusions:

• The developed Bayesian adaptive decision-theoretic design offers an efficient and accurate approach for multi-arm multi-stage clinical trials.
• This methodology enhances the reliability of clinical trial outcomes by optimizing decision-making processes.
• The design is particularly valuable for complex trials involving multiple experimental arms and a control group.