Independent Increments and Group Sequential Tests

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Group sequential tests allow early stopping of clinical trials based on accumulating data.
• Existing methods rely on test statistics possessing the independent increments property.
• This property is often violated in time-to-event trials with non-proportional hazards.

Purpose of the Study:

• To develop new group sequential test statistics that possess the independent increments property.
• To enhance statistical power in clinical trials, particularly when standard assumptions are violated.
• To provide methods compatible with existing group sequential software.

Main Methods:

• Derivation of linear combinations of sequentially computed test statistics.
• Ensuring these linear combinations exhibit the independent increments property.
• Optimizing linear combinations for specific alternative hypotheses (e.g., non-proportional hazards).

Main Results:

• Demonstrated that linear combinations can always achieve the independent increments property, irrespective of the original covariance structure.
• Developed novel sequentially computed test statistics with guaranteed independent increments.
• Showcased increased power against target alternatives compared to methods using original statistics.

Conclusions:

• The proposed methods provide robust group sequential tests applicable even when standard assumptions fail.
• New test statistics support the use of existing software while offering improved power.
• This approach broadens the applicability of early stopping rules in complex clinical trial settings.