Sample size calculation for the combination test under nonproportional hazards.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• The logrank test is standard for sample size calculation in survival trials but assumes proportional hazards (PH).
• The PH assumption is often violated in practice, particularly in immuno-oncology, potentially leading to underpowered studies.
• Combination tests, like the maximum weighted logrank test, offer robust performance across various hazard scenarios.

Purpose of the Study:

• To propose a flexible, simulation-free procedure for sample size calculation using combination tests.
• To address complex clinical trial features such as staggered entry and dropouts.
• To provide a reliable method for sample size determination under non-proportional hazards.

Main Methods:

• Extension of Lakatos' Markov model for sample size calculation.
• Utilized combination tests, including maximum weighted logrank and projection-type tests.
• Evaluated the procedure across diverse hazard scenarios and common statistical tests.

Main Results:

• The proposed method successfully achieved target power for all tested methods in most scenarios.
• Combination tests demonstrated robust performance under both correct and incorrect hazard model specifications.
• The procedure effectively handles complex trial designs with staggered entry and dropouts.

Conclusions:

• The proposed simulation-free procedure is a reliable tool for sample size calculation in clinical trials with survival endpoints, especially under non-proportional hazards.
• Combination tests are highly recommended when hazard-changing patterns are uncertain.
• The method provides practical guidance for sample size determination in real-world clinical trial settings.