Time-to-event analysis with treatment arm selection at interim

Area of Science:

• Clinical Trials
• Biostatistics
• Oncology

Background:

• Adaptive trial designs offer flexibility in clinical research.
• Treatment arm selection is crucial in oncology trials with multiple endpoints.
• Balancing efficacy and type I error control is essential for valid trial conclusions.

Purpose of the Study:

• To apply an adaptive design for treatment arm selection in an oncology trial.
• To combine survival and disease progression data for interim analysis.
• To evaluate statistical methods for maintaining type I error control at final analysis.

Main Methods:

• Utilized an adaptive design for treatment arm selection at an interim analysis.
• Employed Bayesian predictive power to integrate primary (survival) and secondary (disease progression) endpoints.
• Investigated four statistical approaches (Bonferroni, Dunnett-like, conditional error function, combination p-value) for final analysis.

Main Results:

• The study assessed the power and type I error control of different statistical approaches under various conditions.
• Bayesian predictive power was used to combine evidence from survival and progression endpoints for arm selection.
• Frequentist statistical tests were applied to the survival endpoint in the final analysis.

Conclusions:

• Adaptive designs can effectively manage treatment arm selection in oncology trials.
• The chosen statistical methods impact the power and validity of efficacy conclusions.
• Careful consideration of statistical approaches is necessary for robust trial outcomes in adaptive oncology studies.