Flexible Spline Models for Blinded Sample Size Reestimation in Event-Driven Clinical Trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Event-driven trials rely on achieving a specific number of events to maintain statistical power.
• Blinded sample size reestimation (BSSR) adjusts sample size using interim data when planning assumptions are flawed.
• Current BSSR methods often use parametric models for survival function extrapolation, which may be inadequate.

Purpose of the Study:

• To introduce and evaluate a flexible spline-based BSSR method for event-driven trials.
• To compare the performance of the proposed method against traditional parametric approaches.
• To enhance the accuracy of sample size adjustments and improve trial completion time.

Main Methods:

• Developed a BSSR procedure utilizing Royston-Parmar spline models for survival function extrapolation.
• Conducted a simulation study to compare the spline-based method with parametric models.
• Applied the proposed method to a real-world clinical trial in secondary progressive multiple sclerosis.

Main Results:

• The flexible spline-based BSSR method demonstrated robustness, avoiding over- or underestimation of expected events seen with parametric methods.
• Simulations indicated superior performance of the spline-based approach.
• The method's effectiveness was confirmed in a clinical trial application.

Conclusions:

• The proposed flexible spline-based BSSR method offers a more reliable approach for event-driven trials with incorrect planning assumptions.
• This robust method enables more accurate recruitment adjustments, aiding trials in finishing on schedule.
• The spline-based extrapolation provides a valuable alternative to standard parametric models in BSSR.