Determining optimal sample sizes for multistage adaptive randomized clinical trials from an industry perspective...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Health Economics

Background:

• Traditional frequentist methods for sample size determination in clinical trials rely on arbitrary factors.
• Decision-theoretic (Bayesian) approaches, or value of information methods, optimize sample size by maximizing expected net gain (utility minus cost).
• Industry and societal perspectives highlight the potential of multistage designs to enhance expected net gain.

Purpose of the Study:

• Determine optimal sample size for industry-based, multistage adaptive randomized clinical trials using value of information methods.
• Demonstrate the increase in expected net gain achievable with these adaptive designs.
• Evaluate sponsor decisions at each stage: continue, seek approval, or abandon the drug.

Main Methods:

• Develop a model for expected total profit, incorporating per-patient profit, disease incidence, time horizon, trial duration, market share, and regulatory approval probability.
• Extend the model to multistage designs, providing a solution for a two-stage design.
• Illustrate the methodology with a practical example.

Main Results:

• Multistage adaptive designs lead to significant increases in expected net gain compared to traditional approaches.
• The proposed value of information framework effectively guides sample size optimization in adaptive trials.

Conclusions:

• Multistage designs offer substantial gains in expected net gain from a value of information perspective in industry trials.
• While complex, simpler near-optimal solutions exist for multistage designs.
• The methodology assumes sufficient sample size for the central limit theorem to apply, ensuring normal distribution of relevant statistics.