Weighted sum and order statistics methods for dynamic information borrowing in basket trials.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacology

Background:

• Basket trials investigate a single therapy across multiple patient sub-populations under one protocol.
• These trials are crucial for identifying effective treatments in oncology and rare diseases.
• Information borrowing is key to improving statistical efficiency in basket trials.

Purpose of the Study:

• To propose two novel frequentist methods for enhancing statistical efficiency in basket trials.
• To improve treatment effect estimation and optimize multiple testing procedures.

Main Methods:

• Developed a method to minimize mean squared errors in treatment effect estimation by borrowing information.
• Introduced a method utilizing cross-basket information for optimizing multiple testing.

Main Results:

• Proposed methods significantly improve statistical efficiency in basket trials.
• The methods effectively limit family-wise error rate inflation.
• Both methods are compatible with standard statistical models, with or without covariate adjustment.

Conclusions:

• The novel frequentist methods offer substantial improvements for basket trial analysis.
• These methods enhance the ability to identify effective therapies in specific sub-populations.
• The proposed techniques provide robust statistical tools for clinical trial design.