Bayesian adaptive dose-finding studies with delayed responses.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacometrics

Background:

• Bayesian response-adaptive designs enhance dose-finding study efficiency.
• Current interim analysis methods often exclude data from patients who have not completed the study.
• Slow enrollment and long study durations in diseases like diabetes and obesity lead to limited completed data at interim analyses.

Purpose of the Study:

• To propose a novel Bayesian prediction model for interim analyses in clinical trials.
• To incorporate all available patient data, including incomplete and longitudinal data, for improved decision-making.
• To enhance the efficiency and informativeness of adaptive dose-finding studies.

Main Methods:

• Development of a Bayesian prediction model that utilizes all patient data, regardless of completion status.
• Inclusion of methods to handle incomplete longitudinal data, including missing at random (MAR) data.
• Discussion of a utility-function-based decision rule for adaptive trial management.

Main Results:

• Simulations demonstrate the proposed model's superiority over traditional designs in terms of efficiency and decision-making.
• The model effectively incorporates data from both completed and incomplete patient records.
• The method shows improved performance across various simulated trial scenarios.

Conclusions:

• The new Bayesian prediction model offers a significant advancement for adaptive clinical trial design.
• Utilizing all available data, including incomplete records, leads to more informed and efficient interim decisions.
• This approach is particularly beneficial for studies with slow enrollment or long-term treatment effects.