Adaptive group sequential test with changing patient population.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Group sequential tests traditionally assume homogeneous treatment effects over time.
• This assumption is often violated in clinical trials, especially when patient populations change during interim analyses.
• Heterogeneous treatment effects can arise from relaxed inclusion/exclusion criteria, impacting trial validity.

Purpose of the Study:

• To address the issue of inflated Type I error rates in group sequential trials with changing patient populations.
• To develop a statistical method for unbiased inference when treatment effects vary across different patient cohorts.
• To ensure the integrity of statistical analysis in adaptive clinical trial designs.

Main Methods:

• Investigated scenarios where relaxed patient entry criteria at interim analyses lead to heterogeneous treatment effects.
• Utilized linear regression models to analyze data from evolving patient populations.
• Developed a method to make inference on the target population using all available data from changed populations.

Main Results:

• Simulation results demonstrated severe inflation of Type I error rates when no statistical adjustment was made.
• The proposed method, employing linear regression models, effectively corrects for population changes.
• The statistical analysis based on the proposed method yields unbiased inference.

Conclusions:

• Violations of the homogeneity assumption in group sequential trials necessitate statistical adjustments.
• Linear regression models provide a robust framework for unbiased inference in adaptive trials with changing populations.
• The proposed method ensures reliable conclusions despite evolving patient characteristics and treatment effects.