Subgroup detection and sample size calculation with proportional hazards regression for survival data

Area of Science:

• Biostatistics
• Survival Data Analysis
• Clinical Trial Design

Background:

• Identifying patient subgroups with differential treatment effects is crucial for personalized medicine.
• Existing methods may lack robustness or require strong assumptions about baseline models.

Purpose of the Study:

• To propose a novel testing procedure for detecting and estimating subgroups with enhanced treatment effects in survival data.
• To develop a doubly robust score-type test for subgroup identification.

Main Methods:

• A new proportional hazard model incorporating a nonparametric component and a subgroup-treatment-interaction effect (change plane).
• Development of a score-type test robust to misspecification of baseline effects or propensity scores.
• Estimation of subgroup parameters via the supremum of a normalized score statistic.

Main Results:

• The proposed test effectively detects the presence of subgroups with differential treatment effects.
• Asymptotic distributions under null and local alternative hypotheses were established.
• A sample size calculation formula and algorithm for clinical trials were derived.

Conclusions:

• The developed methodology provides a robust framework for subgroup identification in survival analysis.
• The approach facilitates more precise clinical trial design and personalized treatment strategies.
• Demonstrated utility through simulation studies and application to AIDS clinical trial data.