Controlling the type 1 error rate in non-inferiority trials

Area of Science:

• Biostatistics
• Clinical Trials
• Pharmaceutical Research

Background:

• Non-inferiority trials are crucial for evaluating experimental therapies against established controls.
• Existing methods include fixed-margin and synthesis approaches, each with limitations.

Purpose of the Study:

• To introduce a unified approach for non-inferiority trials that encompasses both fixed-margin and synthesis methods.
• To demonstrate how this unified approach controls type 1 error rates, even with violated assumptions.
• To compare the efficiency and power of the unified approach against the fixed-margin method.

Main Methods:

• Development of a unified statistical framework for non-inferiority trial design.
• Parameter selection strategies to control unconditional type 1 error rates.
• Analysis of assay sensitivity and constancy assumption departures.

Main Results:

• The unified approach integrates fixed-margin and synthesis methods as special cases.
• The fixed-margin approach can be highly inefficient.
• The synthesis method, when appropriately chosen, offers equivalent type 1 error control with higher statistical power.

Conclusions:

• The unified approach provides a flexible and powerful framework for non-inferiority trial design.
• Synthesis methods are generally superior to fixed-margin approaches for maximizing power while maintaining error control.
• This research offers improved statistical strategies for experimental therapy evaluation.