Bayesian Analysis of First-Order Markov Models for Autocorrelated Binary Responses

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Longitudinal Data Analysis

Background:

• Patient outcomes in clinical trials are frequently binary and measured asynchronously over time.
• Existing models may not adequately capture autocorrelation in longitudinal binary data or handle asynchronous observations.
• Understanding dose-response relationships is crucial for therapeutic development.

Purpose of the Study:

• To develop a flexible statistical model for analyzing longitudinal binary outcomes in clinical trials.
• To address challenges posed by asynchronous measurements and autocorrelation in patient data.
• To incorporate dose-response relationships and random effects for a comprehensive analysis.

Main Methods:

• A first-order Markov model for binary data was developed to handle autocorrelation.
• Nonhomogeneous models for transition probabilities using B-spline basis functions were proposed for asynchronous data.
• The model was extended to a mixed-effects framework to include individual-specific random effects.
• Bayesian methods with constructed prior distributions were used for estimating non-decreasing dose-response curves.

Main Results:

• The proposed Markov model effectively accounts for autocorrelation in longitudinal binary outcomes.
• The use of B-spline functions enabled modeling of asynchronously observed time points.
• The extended mixed-effects model successfully incorporated random effects, enhancing individual-level analysis.
• Simulations demonstrated superior performance compared to traditional models, and real-world data confirmed practical applicability.

Conclusions:

• The developed Markov model provides a robust framework for analyzing longitudinal binary data in clinical trials with asynchronous observations.
• The model's flexibility allows for the estimation of dose-response curves and incorporation of random effects.
• This approach offers improved accuracy and a more nuanced understanding of patient outcomes in complex trial designs.