Optimizing One-Sample Tests for Proportions in Single- and Two-Stage Oncology Trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Oncology Research

Background:

• Phase II oncology trials frequently use single-arm designs due to cost or rarity of diseases.
• Traditional methods like exact binomial tests and Simon's two-stage designs are often conservative, leading to lower actual Type I error rates than nominal alpha.
• This conservativeness can result in inefficient trial designs with larger sample sizes than necessary.

Purpose of the Study:

• To develop a novel, flexible, and efficient method for early-phase oncology trial design.
• To maintain accurate Type I error control while improving design efficiency.
• To offer a practical alternative to existing conservative trial design methods.

Main Methods:

• A convolution-based statistical method is proposed, combining binomial and simulated normal distributions.
• This method constructs an unbiased estimator for the true response rate (π).
• Theoretical properties are derived, and performance is evaluated against traditional exact tests in one-stage and two-stage designs.

Main Results:

• The proposed method yields more efficient trial designs with reduced sample sizes compared to standard approaches.
• Type I error rates are precisely controlled, matching the nominal alpha level.
• A new two-stage design with interim futility analysis is introduced, demonstrating significant reductions in trial cost and duration.

Conclusions:

• The convolution-based approach provides a flexible and efficient alternative for early-phase oncology trial design.
• It effectively addresses the conservativeness of traditional methods.
• The method offers practical advantages, including reduced resource utilization and shorter study timelines.