Choosing and changing the analysis scale in non-inferiority trials with a binary outcome

Area of Science:

• Biostatistics
• Clinical Trials
• Statistical Methodology

Background:

• Sample size calculations for non-inferiority and equivalence trials are heavily influenced by margin size.
• The choice of analysis scale (risk difference, risk ratio, odds ratio) for binary outcomes significantly impacts statistical power and required sample size, a factor often overlooked.
• Shifting analysis scales during or after a trial can introduce unknown consequences.

Purpose of the Study:

• To outline the sample size implications of selecting different analysis scales at the design stage of non-inferiority trials.
• To evaluate two methods for changing the analysis scale after a trial has begun.
• To assess the impact of scale selection and switching on type I and type II error rates.

Main Methods:

• Comparison of sample size requirements across risk difference, risk ratio, and odds ratio scales.
• Simulation studies to assess type I and type II error rates when changing analysis scales post-commencement.
• Application of methods to the INES trial, a non-inferiority study on fertility treatments.

Main Results:

• Significant differences in required sample size exist between risk difference, risk ratio, and odds ratio scales, with some scales demanding twice the sample size of others.
• Changing the analysis scale mid-trial using anticipated event proportions primarily affects the type II error rate.
• Switching scales post-trial using observed proportions is not recommended as it can compromise the type I error rate.
• The impact of scale choice and switching is more pronounced with larger margins.

Conclusions:

• Researchers must recognize that the choice of analysis scale critically influences type I and type II error rates in non-inferiority trials.
• Awareness of these scale-dependent effects is essential for robust trial design and accurate interpretation of results.