Response-adaptive treatment allocation for clinical studies with ordinal responses

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Modeling

Background:

• Ordinal responses are prevalent in clinical research, often analyzed using proportional odds models.
• Proportional odds models may fail to control Type I error rates when assumptions are violated.
• Latent Weibull models offer superior performance for skewed ordinal data compared to latent normal models.

Purpose of the Study:

• To propose a novel response-adaptive allocation scheme for clinical trials with ordinal outcomes.
• To enhance participant benefit by directing more subjects to superior treatments.
• To ensure the proposed method maintains statistical power for treatment comparisons.

Main Methods:

• Utilizing a doubly adaptive biased coin design for treatment allocation.
• Applying latent Weibull models for analyzing ordinal response data.
• Incorporating a clinical example to demonstrate the procedure's application.

Main Results:

• The doubly adaptive biased coin design effectively allocates more patients to beneficial treatments.
• The proposed allocation scheme preserves adequate statistical power for efficacy assessments.
• Demonstrated feasibility and utility through a relevant clinical case study.

Conclusions:

• The doubly adaptive biased coin design offers an ethical and efficient approach for clinical trials with ordinal data.
• This method improves upon traditional balanced designs by prioritizing participant well-being.
• The latent Weibull model combined with adaptive allocation provides a robust framework for ordinal outcome analysis.