Sample size estimation for the averted events ratio.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• The averted events ratio (AER) is a novel estimand for non-inferiority trials with time-to-event outcomes.
• Unlike traditional measures, AER focuses on averted events rather than observed events, requiring assumptions about background event rates or control treatment effectiveness.
• This study develops sample size formulas for AER and compares them to the conventional rate ratio method.

Purpose of the Study:

• To develop and present sample size formulae for non-inferiority active-control prevention trials using the averted events ratio (AER) as the primary estimand.
• To compare the sample size requirements of AER-based methods with the conventional 95-95 method.
• To evaluate the impact of different assumptions (counterfactual placebo incidence vs. control treatment effectiveness) on sample size.

Main Methods:

• Sample size is expressed using expected events and person-years follow-up.
• Formulae are derived using Wald confidence intervals on a logarithmic scale.
• Sample size calculations consider background event rate, control effectiveness, non-inferiority margin, confidence limits, and statistical power.

Main Results:

• The AER based on counterfactual placebo incidence yields the smallest sample sizes, with reductions of 2.6- to 11.9-fold compared to the 95-95 method for 50-80% control effectiveness.
• The AER based on control treatment effectiveness also reduces sample size (1.5- to 6.4-fold reductions).
• Sample size is highly sensitive to the non-inferiority margin, with increases of 1.8- to 2.5-fold when the margin increases from 50% to 80%.

Conclusions:

• The AER offers advantages in interpretability and allows for smaller, more cost-effective non-inferiority trials.
• Deriving AER via counterfactual placebo incidence is preferred when practicable.
• These findings support the adoption of AER in active-control non-inferiority trial design.