Continuous-time Markov modelling of flexible-dose depression trials

Area of Science:

• Pharmacometrics
• Systems Biology
• Biostatistics

Background:

• Longitudinal data analysis is crucial for understanding disease progression and treatment effects.
• Limited knowledge of disease mechanisms complicates traditional modeling approaches.
• Accurate modeling is essential for clinical trial design and interpretation, especially with flexible dosing.

Purpose of the Study:

• To develop a systematic methodology for modeling longitudinal data in situations with minimal or no prior knowledge of disease pathophysiology.
• To propose a continuous-time stochastic approach within a system-theoretic framework to model patient health states and clinical endpoints.
• To validate the proposed modeling methodology using data from a Phase II depression trial with flexible dosing.

Main Methods:

• A system-theoretic paradigm was adopted to develop a population response model.
• A continuous-time stochastic approach was employed, defining patient health state by clinical score and its time-derivative.
• Dose escalations were modeled as instantaneous perturbations, and parameter estimation used the empirical Bayes method.

Main Results:

• The proposed modeling approach was validated on experimental data from a Phase II depression trial (placebo and drug arms).
• An integrated Wiener process model demonstrated capability in capturing individual responses.
• A stable Markovian model provided a superior description of inter-individual variability compared to simpler models.

Conclusions:

• The developed methodology offers a robust framework for modeling longitudinal data, particularly when disease mechanisms are not fully understood.
• The continuous-time stochastic approach effectively handles complex trial data, including flexible dosing adjustments.
• The study highlights the utility of system-theoretic models and Markovian processes in characterizing disease progression and inter-individual variability in clinical settings.