Bayesian population modeling of drug dosing adherence.

Area of Science:

• Pharmacometrics
• Population modeling
• Bayesian statistics

Background:

• Medication adherence significantly impacts drug concentration and treatment efficacy.
• Understanding variations in adherence, dose timing, overdosing, and persistence is crucial for effective pharmacotherapy.
• Existing models often lack integration and a population-level perspective for these adherence behaviors.

Purpose of the Study:

• To develop an integrated population model describing variations in adherence, dose-timing deviations, overdosing, and persistence.
• To extend an individual adherence model to the population level using a Bayesian approach.
• To compare and identify the most suitable model for population adherence analysis.

Main Methods:

• A hybrid Markov chain-von Mises method was adapted for population modeling via a Bayesian framework.
• Four integrated population models were formulated and compared, utilizing Markov chain-Monte Carlo algorithms.
• The models were validated using medication event monitoring system data from 207 hypertension patients.

Main Results:

• All four Bayesian models exhibited good convergence and mixing properties.
• Adherence, dose-timing deviations, overdosing, and persistence distributions were non-normal and diverse.
• A model incorporating a cooperativity term and hyperbolic parameterization for transition probabilities was preferred.
• Simulated distributions from the preferred model closely matched observed patient data.

Conclusions:

• The developed Bayesian population model offers a parsimonious and integrated description of medication adherence.
• This model can be applied to clinical trial simulations and pharmacokinetic-pharmacodynamic modeling.
• The findings highlight the non-normal and diverse nature of adherence-related behaviors in patient populations.