Event-driven planning of two-armed trials with a binary endpoint

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Power Analysis

Background:

• Sample size calculation for clinical trials with binary endpoints often relies on event probability, which can be uncertain.
• Survival trials are powered by the number of events, which is less sensitive to unknown parameters than sample size.
• This study investigates the relative stability of event counts versus sample size in two-armed randomized trials with binary outcomes.

Purpose of the Study:

• To quantify the relative stability of the number of events compared to sample size for binary endpoint trials.
• To explore the enhancement of adaptive trial designs using this relative stability.
• To evaluate the potential benefits of a simple event-driven strategy in such settings.

Main Methods:

• Utilized sample size formulas to assess the stability of event numbers versus sample size for relative risk, odds ratio, and risk difference.
• Conducted simulations to evaluate an event-driven design under conditions of relative stability.
• Assessed type I error rate and power using various analysis methods and trial halting strategies.

Main Results:

• The number of events is at least three times more stable than sample size for relative risk (event probability < 1/3) and odds ratio (event probability < 0.20).
• This stability is independent of error rates and treatment effect magnitude.
• Simulations of event-driven designs showed that while asymptotic methods may inflate type I error, other approaches demonstrate favorable operating characteristics.

Conclusions:

• In trials with moderately low event probabilities, focusing on the number of events can aid trial planning and futility assessment.
• This approach may facilitate the development of simple, feasible, and appealing event-driven designs for binary endpoints.
• Adopting an event-centric perspective can streamline clinical trial planning and execution.