A comparison of methods for sample size estimation for non-inferiority studies with binary outcomes

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Pharmaceutical Research

Background:

• Ethical considerations increasingly preclude placebo-controlled trials when effective treatments exist.
• Active-controlled trials are essential for comparing new interventions against established standards of care.
• Non-inferiority trials aim to demonstrate that a new treatment is not unacceptably worse than an existing one.

Purpose of the Study:

• To review and compare methodologies for calculating sample sizes in non-inferiority trials.
• To provide guidance on appropriate sample size calculation methods.
• To assess the sensitivity of sample size to assumptions, particularly for binary outcomes.

Main Methods:

• Review of existing sample size calculation methodologies for non-inferiority trials.
• Comparison of different methods through simulation studies.
• Analysis of sample size sensitivity to control response rates in binary outcome trials.

Main Results:

• The simplest method for sample size calculation, using anticipated response, aligns well with simulation results.
• Sample size calculations for binary outcomes with high expected response rates are highly sensitive to assumptions about the control group's response.
• Bayesian methods offer a robust approach for sample size assessment in these scenarios.

Conclusions:

• The simplest sample size calculation method is generally reliable.
• Sensitivity analyses are crucial when designing non-inferiority trials with binary outcomes and high response rates.
• Adoption of Bayesian methods is recommended for robust sample size determination in non-inferiority trials.