Extending the classification approach for comparing two active treatment arms to binary and time-to-event outcomes

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacoeconomics

Background:

• Regulatory trials comparing active treatments require prespecified hypotheses (noninferiority/superiority).
• Limited comparative efficacy data poses challenges for selecting the best treatment.
• Existing methods often rely on hypothesis testing, which may be unsuitable when prior information is scarce.

Purpose of the Study:

• To extend the classification methodology of Qu et al. for comparing two active treatments.
• To provide an estimation-based approach for treatment comparison with binary and time-to-event outcomes.
• To offer a framework for identifying the best treatment when comparative efficacy is uncertain.

Main Methods:

• Extends Qu et al.'s classification approach for binary and time-to-event data.
• Utilizes an estimation-based framework instead of traditional hypothesis testing.
• Divides the decision space into eight outcomes based on non-inferiority margins.
• Draws conclusions from the confidence interval of relative risk or hazard ratio.

Main Results:

• The proposed method theoretically controls the misclassification rate at a specified level.
• Simulations demonstrate the method's practical application and reliability.
• The approach was illustrated using data from a Phase 3 nonsmall cell lung cancer study.

Conclusions:

• This estimation-based classification method offers a valuable alternative for comparing active treatments, especially with limited prior efficacy information.
• The methodology provides a structured way to classify comparative treatment outcomes.
• It aids in identifying the superior treatment by analyzing confidence intervals for relative risk or hazard ratios.