A frequentist design for basket trials using adaptive lasso

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Basket trials accelerate drug development by testing therapies across diverse patient populations.
• Traditional analysis treats each basket independently, potentially missing cross-basket treatment effects.
• Bayesian dynamic borrowing methods share data but risk type 1 error inflation.

Purpose of the Study:

• To propose a novel frequentist dynamic borrowing method for basket trials.
• To evaluate the performance of adaptive lasso in basket trial analysis.
• To offer a more implementable and faster alternative to existing methods.

Main Methods:

• A frequentist dynamic borrowing approach using adaptive lasso was developed.
• Simulation studies were conducted to compare adaptive lasso with existing methods.
• The method was assessed for power and type 1 error rates.

Main Results:

• Adaptive lasso demonstrated comparable statistical power to Bayesian dynamic borrowing methods.
• The proposed frequentist approach maintained acceptable type 1 error rates.
• Adaptive lasso proved easier to implement and faster than Bayesian alternatives.

Conclusions:

• Adaptive lasso provides an effective frequentist strategy for basket trials.
• This method offers a practical and efficient alternative for analyzing multi-population drug studies.
• The flexibility of adaptive lasso allows for extension to various trial designs and endpoints.