Censoring-robust estimation in fixed sample time-to-event clinical trials with adaptive randomization

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Survival Analysis

Background:

• Adaptive randomization dynamically adjusts treatment allocation probabilities during clinical trials.
• Its impact on estimating treatment efficacy in time-to-event trials with time-varying effects is not fully understood.
• Existing methods may be unreliable when treatment effects change over time.

Purpose of the Study:

• To investigate the effects of adaptive randomization on estimating marginal hazard ratios in time-to-event trials.
• To develop and validate a robust statistical method to address potential biases introduced by adaptive randomization.
• To apply the proposed method to real-world clinical trial data.

Main Methods:

• Analytical derivation showing adaptive randomization alters censoring patterns.
• Monte Carlo simulations to assess bias in the Cox proportional hazards estimator.
• Development of a censoring-robust estimator using reweighted partial likelihood scores.
• Derivation of asymptotic properties and finite sample evaluation of the proposed estimator.

Main Results:

• Adaptive randomization demonstrably alters censoring patterns in time-to-event trials.
• The standard Cox proportional hazards estimator can produce biased results under adaptive randomization.
• The proposed censoring-robust estimator effectively corrects for these biases.
• The method's performance was validated through simulations and application to a real AIDS clinical trial.

Conclusions:

• Adaptive randomization requires careful consideration in time-to-event trial analysis due to potential bias.
• The proposed robust estimator provides a reliable method for estimating treatment efficacy.
• This approach enhances the accuracy of survival analysis in adaptive clinical trials.