Bayesian phase II adaptive randomization by jointly modeling efficacy and toxicity as time-to-event outcomes.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Drug Development

Background:

• Phase II trials aim to assess treatment efficacy and monitor adverse effects.
• Current adaptive randomization (AR) methods often model toxicity as a binary endpoint, failing to capture evolving toxicity profiles.
• Existing approaches use unobservable random effects (frailty) to link efficacy and toxicity.

Purpose of the Study:

• To propose a new Bayesian adaptive randomization (BAR) procedure for Phase II clinical trials.
• To jointly model efficacy and toxicity as time-to-event (TTE) outcomes using a covariate-adjusted efficacy-toxicity ratio (ETR) index.
• To introduce early stopping rules for toxicity and futility to discontinue inferior treatments sooner.

Main Methods:

• Developed a novel Bayesian adaptive randomization (BAR) procedure.
• Modeled efficacy and toxicity as time-to-event (TTE) outcomes.
• Incorporated a covariate-adjusted efficacy-toxicity ratio (ETR) index.
• Proposed early stopping rules for toxicity and futility.

Main Results:

• The proposed BAR procedure demonstrated superior identification of treatment toxicity differences compared to existing methods.
• Simulations indicated that the new BAR approach can better allocate patients to superior treatment arms.
• The method effectively handles evolving toxicity profiles over time.

Conclusions:

• The proposed Bayesian adaptive randomization (BAR) procedure offers an improved approach for Phase II clinical trials.
• Jointly modeling efficacy and toxicity as time-to-event outcomes enhances treatment allocation and patient safety.
• Early stopping rules contribute to greater trial efficiency by discontinuing ineffective treatments promptly.