A Minimal Continuous-Time Markov Pharmacometric Model | JoVE Visualize

Area of Science:

Biostatistics
Statistical Modeling
Data Analysis

Background:

Analyzing ordered categorical data with Markov properties often requires complex models like discrete-time Markov models (DTMM) or standard continuous-time Markov models (CTMM).
Existing models may face limitations with non-uniform observation times or a large number of categories.
There is a need for a more parsimonious yet effective model for such data.

Purpose of the Study:

To introduce and evaluate the minimal continuous-time Markov model (mCTMM) as an alternative for analyzing ordered categorical data with Markov properties.
To demonstrate the mCTMM's ability to simplify Markov property representation and integrate with proportional odds (PO) models for steady-state probabilities.
To assess the mCTMM's performance against existing models using real-world patient data.

Main Methods:

Developed the minimal CTMM (mCTMM) by reparameterizing CTMM, assuming transition rates are state-independent.
Expressed Markov property via mean equilibration time and steady-state probabilities via a proportional odds (PO) model.
Compared mCTMM performance against PO models and published Markov models using fatigue and pain score datasets.

Main Results:

mCTMM provided a better fit to the data than the PO model (ignoring Markov features) across all examples.
mCTMM accurately predicted average transitions and maximum achieved scores, outperforming a count model for Likert data.
While less flexible than DTMM, mCTMM required fewer parameters and demonstrated applicability with non-uniform time intervals and numerous categories.

Conclusions:

The mCTMM offers a valuable, parsimonious approach for analyzing ordered categorical data with Markov properties, especially when standard models are challenging to implement.
This model facilitates the exploration of predictive factors like drug exposure and covariates while accounting for underlying Markovian dynamics.
mCTMM serves as a practical alternative for complex datasets where DTMM or standard CTMM may not be suitable.