Modified Goldilocks Design with strict type I error control in confirmatory clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Bayesian Statistics

Background:

• Goldilocks Design (GD) is a Bayesian adaptive design that uses predictive probability for sample size adaptation.
• GD requires extensive simulations to control Type I error for specific null space subsets, limiting its use in confirmatory trials.

Purpose of the Study:

• To propose a Modified Goldilocks Design (MGD) for adaptive clinical trials.
• To analytically control Type I error across the entire null space, enhancing applicability for confirmatory trials.

Main Methods:

• The MGD applies the conditional invariance principle.
• It utilizes a combination test approach on p-values derived from independent cohorts.
• This method analytically controls Type I error without extensive pre-study simulations.

Main Results:

• The MGD analytically controls Type I error across the entire null space.
• Simulation studies demonstrate that MGD maintains statistical power comparable to the original GD.
• The design was successfully applied to a time-to-event oncology trial.

Conclusions:

• The Modified Goldilocks Design provides an analytical method for Type I error control in Bayesian adaptive trials.
• MGD is a viable alternative to GD for confirmatory trials, offering improved Type I error control with similar power.
• This approach simplifies the design process by reducing reliance on extensive simulations.