On the Conditional Power in Survival Time Analysis Considering Cure Fractions

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Survival Analysis

Background:

• Conditional power is vital for interim clinical trial analyses, guiding decisions on trial continuation or futility.
• Simple survival models struggle with cure fractions, leading to inaccurate conditional power estimates.
• Non-mixture models offer a framework to accurately account for cure fractions in survival data.

Purpose of the Study:

• To derive conditional power functions for non-mixture survival models.
• To implement these functions in an R package for practical application.
• To compare the performance of non-mixture models against simple models using clinical trial data.

Main Methods:

• Derivation of conditional power functions for non-mixture exponential, Weibull, and Gamma models.
• Implementation of derived formulae in the R package 'CP'.
• Application and comparison using a real-world clinical trial dataset.

Main Results:

• Conditional power functions for non-mixture models were successfully derived.
• The R package 'CP' was developed to facilitate these calculations.
• Results demonstrated differences in conditional power estimates compared to the simple exponential model, particularly when cure fractions are present.

Conclusions:

• Non-mixture models provide a more accurate approach to conditional power calculation in the presence of cure fractions.
• The developed R package 'CP' can aid biostatisticians in making informed decisions during clinical trial interim analyses.
• Accurate conditional power estimation using appropriate models is essential for efficient and ethical clinical trial conduct.