Power Calculation for Non-inferiority Test Based on Linear Combination of Two Correlated Binary Endpoints

Area of Science:

• Biostatistics
• Clinical Trials
• Cancer Research

Background:

• Multiple correlated primary endpoints are common in cancer research.
• Independent evaluation of endpoints in sample size calculations can be overly conservative.
• Accounting for endpoint correlation is crucial for efficient study design.

Purpose of the Study:

• To develop and evaluate sample size calculation methods for correlated endpoints in non-inferiority trials.
• To propose a power calculation approach using a linear combination of two correlated clinical outcomes.
• To investigate the impact of various factors on sample size requirements.

Main Methods:

• Derivation of sample size formulae for non-inferiority hypothesis testing.
• Utilizing a linear combination of two correlated clinical outcomes.
• Theoretical evaluation and numerical simulations to assess results.

Main Results:

• Required sample size is sensitive to endpoint correlation (ρ), event probabilities (p1, p2), and non-inferiority margin (δZ).
• Higher correlation and equal event rates increase sample size; larger non-inferiority margins decrease it.
• Weights assigned to composite endpoints significantly affect variance and sample size; balanced weights reduce sample size.

Conclusions:

• A linear combination approach for correlated endpoints offers a more efficient alternative to independent analysis for sample size calculations.
• Study design parameters like correlation, event rates, non-inferiority margin, and weighting significantly influence power and sample size.
• Careful consideration of these factors is essential for optimizing clinical trial sample sizes in cancer research.