Adaptive Optimal Designs for Dose-Finding Studies with Time-to-Event Outcomes

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Dose-finding studies are crucial for determining optimal drug dosages.
• Censored time-to-event data, common in clinical trials, presents analytical challenges.
• Weibull models are frequently used for time-to-event data analysis.

Purpose of the Study:

• To investigate optimal design strategies for dose-finding studies with censored Weibull outcomes.
• To compare the efficiency of adaptive D-optimal designs against traditional fixed designs.
• To evaluate the impact of censoring and dose-response models on design efficiency.

Main Methods:

• Utilized locally D-optimal designs for a quadratic dose-response model.
• Employed two-stage adaptive D-optimal designs with maximum likelihood estimation (MLE) model updating.
• Conducted simulations across various dose-response scenarios and censoring levels.

Main Results:

• Adaptive optimal designs demonstrated nearly equivalent efficiency to locally D-optimal designs.
• Equal allocation designs showed significant inefficiency with high censoring and increasing Weibull hazard.
• Adaptive designs with early stopping showed potential for reduced sample sizes without compromising precision.

Conclusions:

• Adaptive D-optimal designs offer a robust and efficient approach for dose-finding studies with censored time-to-event data.
• Early stopping in adaptive trials can optimize resource allocation and maintain statistical power.
• Fixed equal allocation designs may be suboptimal in scenarios with substantial data censoring.