Benchmarking non-inferiority design and analysis through the Z-score

Non-inferiority (NI) design and analyses are frequently employed to provide alternative therapy options to patients, ensuring these therapies are not significantly worse in terms of safety and effectiveness compared to the standard therapy. However, unlike superiority analysis, non-inferiority analysis faces the inherent challenge of a potentially irrelevant NI margin. It is not uncommon for the observed control performance to differ significantly from the assumed one, which can render the NI margin irrelevant. The purpose of this paper is to utilize Z-score to investigate this issue.

This paper proposes using the Z-score to evaluate the NI design and connect the NI design and analysis. This approach provides a benchmark tool that even when actual outcomes differ significantly from assumptions, the NI margin can be moderated to balance clinical relevance and statistical practicality.

Illustrative calibrations and case studies demonstrate the feasibility of this approach, facilitating the transition from NI design to analysis.

The Z-score can be used to assess the quality of the NI design and serve as a bridge between NI design and analysis.