Incorporating historical two-arm data in clinical trials with binary outcome: A practical approach

Area of Science:

• Clinical Trials Methodology
• Bayesian Statistics
• Frequentist Statistics

Background:

• Historical data incorporation can enhance clinical trial feasibility.
• A key challenge is controlling the inflated Type I error rate.
• Existing methods lack clear recommendations for single historical trial data.

Purpose of the Study:

• To propose a frequentist framework for incorporating historical data from two-armed trials with binary outcomes.
• To control the Type I error rate while utilizing historical information.
• To explore potential reductions in required sample size.

Main Methods:

• Adaptation of the Bayesian power prior approach to a frequentist setting.
• Introduction of a parameter (δ) to control the amount of borrowed historical data.
• Development of methods to determine δ for controlling Type I error and reducing sample size.

Main Results:

• A frequentist framework is proposed for incorporating historical data from both arms of two-armed trials.
• For any trial scenario, a suitable δ can be found to maintain the Type I error rate below the significance level.
• The method allows for increased statistical power and reduced sample size compared to trials without historical data borrowing.

Conclusions:

• The proposed frequentist power prior framework effectively controls Type I error rates when incorporating single historical trial data.
• This approach offers flexibility in borrowing historical information, leading to potential power gains and sample size reductions.
• The methodology provides a valuable alternative to Bayesian approaches for integrating historical data in clinical trial design.