Bayesian randomized basket trial design: a case study from the ultra-rare invasive mold infections

Area of Science:

• Clinical Trials
• Biostatistics
• Infectious Diseases

Background:

• Invasive mold infections (IMIs) are rare, life-threatening fungal diseases.
• Conducting well-powered randomized noninferiority trials for IMIs is challenging due to low patient accrual.
• Heterogeneity among mold types further complicates study design and treatment interpretation.

Purpose of the Study:

• To develop efficient statistical methods for randomized basket trials in non-oncologic rare diseases like IMIs.
• To enhance the efficiency and precision of treatment effect estimates across different mold types.
• To improve the ethical feasibility of rare disease trials by reducing control arm patient numbers.

Main Methods:

• Proposed a robust borrowing strategy for randomized basket trials.
• Incorporated borrowing treatment effects across mold types while accounting for heterogeneity.
• Augmented control arms using external data.

Main Results:

• The proposed approach significantly increased statistical power and precision for treatment effect estimates.
• Demonstrated maintenance of the family-wise type I error rate at an acceptable level.
• Showcased improvements over traditional pooling methods for rare disease inference.

Conclusions:

• The developed statistical approach offers a substantial improvement for rare disease clinical trials, particularly for invasive mold infections.
• This method enhances trial efficiency, precision, and ethical considerations.
• Applicable to other rare diseases facing similar patient accrual and ethical challenges.