Covariate-Adjusted Group Sequential Comparisons of Survival Probabilities.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Confirmatory clinical trials frequently analyze survival outcomes, necessitating interim analyses for efficacy and futility.
• Standard methods like the log rank test and Cox regression are sensitive to violations of the proportional hazards (PH) assumption, potentially reducing statistical power.
• Treatment mechanisms, such as immunotherapy versus chemotherapy, can lead to expected PH violations in cancer trials.

Purpose of the Study:

• To propose novel group sequential tests for comparing survival probabilities with covariate adjustment.
• To enable interim analyses that accommodate non-proportional hazards (non-PH) and provide interpretable treatment effect measures.
• To facilitate Type I error control and sample size determination in sequential survival studies.

Main Methods:

• Development of group sequential tests for survival data incorporating covariate adjustment.
• Utilizing the asymptotic independent increments structure of test statistics for simplified critical value specification.
• Application of methods to compare survival probabilities in the presence of non-PH.

Main Results:

• Simulations confirm that the proposed tests maintain targeted Type I error rates and power.
• The tests demonstrate robustness to violations of the proportional hazards assumption and covariate influence.
• The methodology was successfully illustrated using data from the Blood and Marrow Transplant Clinical Trials Network 1101 trial.

Conclusions:

• The proposed group sequential tests offer a robust approach for analyzing survival data in clinical trials with potential non-PH.
• These methods provide clinically meaningful summary measures and maintain statistical integrity during interim analyses.
• The approach enhances the reliability of survival outcome comparisons in diverse treatment settings.