Sample size re-estimation in a breast cancer trial

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Randomized breast cancer trials require accurate sample size calculations.
• Initial estimates of failure probabilities can be overestimated, potentially reducing trial power.
• Existing sample size re-estimation methods often focus on binary or normal outcomes, necessitating new approaches for time-to-event data.

Purpose of the Study:

• To develop and apply a blinded sample size re-estimation method for time-to-event trials.
• To address the issue of potentially overestimated failure rates impacting trial power.
• To combine current trial data with prior information or parametric models for re-estimation.

Main Methods:

• Developed a blinded sample size re-estimation technique for time-to-event data.
• Utilized current blinded trial data and prior study information to re-estimate failure probabilities.
• Employed bootstrap resampling to assess uncertainty in re-estimated sample sizes.

Main Results:

• Re-estimation using data from 278 patients showed minimal change in required sample size.
• The developed method preserves the type I error rate.
• When initial failure probability assumptions are correct, the median sample size increase is zero.

Conclusions:

• Blinded sample size re-estimation is valuable for lengthy accrual period trials.
• Prior knowledge of survival distribution or prior data is essential for accurate re-estimation.
• The method generally results in minimal or no sample size increase when initial assumptions are valid.