Recurrent events modeling based on a reflected Brownian motion with application to hypoglycemia.

Area of Science:

• Biostatistics
• Endocrinology
• Mathematical Modeling

Background:

• Type 2 diabetes management requires diligent blood sugar monitoring.
• Hypoglycemia is a common adverse event despite glucose-controlling treatments.
• Patients readily perceive hypoglycemia symptoms, aiding its observation.

Purpose of the Study:

• To develop a mathematical model for analyzing hypoglycemic events in type 2 diabetes.
• To quantify glucose variability, patient heterogeneity, and risk factor impact.
• To provide a framework for improved diabetes self-management and treatment strategies.

Main Methods:

• Modeled hypoglycemic events as lower boundary crossings of reflected Brownian motion.
• Incorporated patient covariates and frailty into volatility and upper reflection barriers.
• Utilized a Bayesian framework with Markov chain Monte Carlo for inference.
• Employed deviance information criterion and pseudo-marginal likelihood for model selection.

Main Results:

• The proposed model demonstrated adequate fit to patient data.
• The model successfully generated data similar to observed hypoglycemic events.
• Simulation studies validated the methodology's effectiveness.
• Analysis of DURABLE trial data provided significant insights into glucose dynamics.

Conclusions:

• The Brownian motion framework offers a robust method for analyzing glucose variability in type 2 diabetes.
• The model effectively captures patient heterogeneity and the influence of risk factors.
• This approach provides valuable insights for clinical practice and diabetes self-management.