Implementing a Bayesian approach using Stan with Torsten: Population pharmacokinetics analysis of somatrogon

Area of Science:

• Pharmacometrics
• Computational Statistics
• Drug Development

Background:

• Population pharmacokinetic (PK) modeling is crucial for understanding drug behavior in diverse patient groups.
• Fully Bayesian approaches are underutilized in population PK modeling despite their potential benefits.

Purpose of the Study:

• To evaluate the utility of Stan with R and Torsten for population PK modeling of somatrogon.
• To assess the impact of different prior sets and parameterization strategies on model performance and computational efficiency.

Main Methods:

• Bayesian inference using Stan for Markov Chain Monte Carlo (MCMC) sampling.
• Population PK modeling of somatrogon, a long-acting growth hormone.
• Evaluation of three prior sets (weakly, moderately, and very informative) and centered vs. non-centered parameterization.

Main Results:

• All evaluated prior sets demonstrated good performance with well-mixed MCMC chains.
• Posterior predictions accurately covered observed data for both existing and new individuals.
• Non-centered parameterization improved estimation time, though computational intensity remained a factor.

Conclusions:

• Bayesian approaches with Stan and Torsten are viable for population PK analysis, especially with sufficient computational resources.
• These methods are particularly useful for special populations, small datasets, and complex model structures.
• The study highlights the practical application of Bayesian inference in modern pharmacometrics.