A geometrically intuitive framework for the synthesis method in non-inferiority trials

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Pharmaceutical Research

Background:

• Non-inferiority (NI) trials are crucial for evaluating new treatments against established ones.
• The fixed margin method (FMM) is the classical approach, using a pre-defined NI margin (M).
• The synthesis method (SM) offers greater efficiency by incorporating historical data but is less commonly used.

Purpose of the Study:

• To address perceived limitations of the synthesis method (SM) in non-inferiority (NI) trials.
• To demonstrate a novel parameterization for SM that allows for pre-specification of the NI margin.
• To provide a clear visualization and analytical framework for both FMM and SM.

Main Methods:

• The study introduces a bivariate space parameterization for NI trial analysis, plotting Δ_NI (observed active-controlled effect) against Δ_H (historical control vs. placebo effect).
• The NI margin is defined in the protocol by the line Δ_NI + (1-δ)Δ_H = 0, where δ is the minimum acceptable efficacy retention fraction.
• Both FMM (confidence segment) and SM (confidence ellipse) are visualized in this bivariate plane, allowing for direct comparison against the defined NI margin line.

Main Results:

• The bivariate parameterization effectively overcomes the limitations of pre-specifying an NI margin and presenting results in SM.
• Visualizations intuitively represent FMM as a horizontal confidence segment and SM as a confidence ellipse within the bivariate space.
• Formal mathematical demonstration confirms that comparing these regions to the NI margin line is equivalent to conventional Z-tests for NI evaluation under both FMM and SM.

Conclusions:

• The bivariate space approach provides a unified and intuitive framework for conducting and interpreting non-inferiority trials using both fixed margin and synthesis methods.
• This method enhances the practical application of the more efficient synthesis method by addressing its previous limitations.
• The findings support the wider adoption of the synthesis method in clinical trial design and analysis for improved efficiency.