Optimizing disease progression study designs for drug effect discrimination.

Area of Science:

• Clinical Trials
• Biostatistics
• Drug Development

Background:

• Optimizing clinical trial design is crucial for efficiently detecting drug effects.
• Traditional methods often prioritize parameter precision over direct statistical power optimization.
• Existing power prediction methods, like the Wald statistic, can be inaccurate.

Purpose of the Study:

• To investigate direct optimization of clinical trial designs for statistical power.
• To compare power-optimal designs with those focused on parameter precision.
• To develop a new statistic for explicit power optimization in clinical trial design.

Main Methods:

• Developed an improved statistic based on the Wald approximation for predicting statistical power.
• Compared the new statistic and classical Wald statistic against model-based power via simulations.
• Maximized the new statistic to determine power-optimal study designs.
• Evaluated empirical power of optimized designs against parameter precision-focused designs.

Main Results:

• The new statistic showed excellent agreement with empirically determined power, unlike the classical Wald statistic which over-predicted power by up to 90%.
• Power-optimal designs achieved equal or up to 20% higher statistical power compared to traditional designs.
• The method successfully minimized participant numbers while achieving 80% power, confirmed by simulations.

Conclusions:

• A novel statistic enables direct optimization of clinical trial designs for statistical power.
• This approach leads to more powerful and efficient trial designs compared to traditional methods.
• The developed method can optimize both study design and size for targeted statistical power.