Determining the duration of comparative clinical trials while allowing for cure

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Estimating accrual length in randomized survival trials is crucial for resource allocation.
• Current methods often assume exponential failure, which may not fit data with cure fractions.
• Pediatric cancers frequently exhibit cure fractions, making standard assumptions potentially misleading.

Purpose of the Study:

• To propose improved methods for estimating accrual duration in randomized survival trials.
• To address limitations of the exponential failure assumption when cure fractions are present.
• To introduce flexible failure models for survival data with non-constant hazard rates.

Main Methods:

• Utilizing a general formulation for accrual length estimation.
• Incorporating failure models with hazard functions that decrease to zero.
• Applying two specific models to analyze data from Childrens Cancer Study Group (CCSG) trials.

Main Results:

• The proposed models provide a good fit for survival data with cure fractions.
• The choice of failure model significantly impacts the estimated required trial duration.
• The exponential failure assumption can lead to inaccurate accrual length estimations in relevant scenarios.

Conclusions:

• Flexible failure models are essential for accurate accrual length estimation in trials with cure fractions.
• The proposed methodology enhances the reliability of planning for randomized survival trials, particularly in pediatric oncology.
• Accurate modeling of failure times is critical for efficient clinical trial design and duration determination.