[Subgroup identification based on an accelerated failure time model combined with adaptive elastic net]

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Identifying patient subgroups with differential treatment effects is crucial for personalized medicine.
• Accelerated Failure Time (AFT) models offer a flexible framework for survival data analysis.
• Traditional methods may lack power in detecting treatment effects within specific subgroups.

Purpose of the Study:

• To propose a novel strategy for identifying subgroups with treatment effects using AFT models.
• To evaluate the performance of a penalized AFT model and a two-stage adaptive design for subgroup analysis.
• To compare the proposed methods against traditional approaches in clinical trial settings.

Main Methods:

• Applied adaptive elastic net to an AFT model, incorporating covariate-treatment interactions.
• Utilized a likelihood-based change-point algorithm for subgroup classification.
• Adopted a two-stage adaptive design to validate treatment effects in identified subgroups.

Main Results:

• The penalized AFT model outperformed univariate models in challenging scenarios (small sample size, high censoring, small subgroup size).
• The two-stage adaptive design demonstrated improved power for detecting treatment effects where subgroup effects exist.
• The adaptive design maintained well-controlled Type I error rates.

Conclusions:

• The penalized AFT model with main covariate effects is advantageous for subgroup identification in clinical trial survival data.
• The two-stage adaptive design offers superior performance in evaluating treatment effects when subgroup effects are present, compared to traditional designs.