Individualized Multi-Day T1D Simulation for Sensor-Driven Dosing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing simulation models for Type 1 Diabetes (T1D) management fail to accurately simulate insulin therapies using either Self-Monitored Blood Glucose (SMBG) or Continuous Glucose Monitoring (CGM) due to the absence of realistic patient decision-making models, intra-individual variability, and errors in diabetes management, limiting their effectiveness in preventing hyper/hypoglycemic events.
Innovation Solution
A comprehensive mathematical model that simulates T1D patient decision-making, incorporating SMBG and CGM measurement errors, and accounts for patient-specific behaviors such as miscalculations and insulin administration errors, allowing multi-day simulations and realistic glucose/insulin dynamics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing simulation models are used without patient decision-making components, then the model structure is simpler, but the accuracy in simulating insulin therapies and predicting hyper/hypoglycemic events deteriorates
Solution Approach 1:
The simulation model is divided into distinct functional modules: a glucose-insulin dynamic model, a patient decision-making model, and a measurement error model. Each module handles specific aspects of diabetes management, allowing the complex system to be constructed from manageable components that can be independently validated and optimized.
Solution Approach 2:
The patient decision-making model is nested within the simulation framework, with the decision-making logic embedded in the glucose-insulin dynamic model. Measurement error models are nested within the measurement components, creating a hierarchical structure where simpler models are contained within more complex ones.
2Reliability
If CGM is used for insulin dosing, then glycemic control is improved, but safety concerns and lack of definitive evidence increase
Solution Approach 1:
The simulation model incorporates feedback loops where CGM measurements are continuously used to adjust insulin dosing recommendations. The model compares CGM-driven dosing against SMBG-driven dosing and against optimal control strategies, providing feedback on the safety and effectiveness of CGM usage in virtual patient populations.
Solution Approach 2:
The model pre-simulates potential adverse events and safety risks associated with CGM-driven insulin dosing before they occur in clinical practice. By running multiple virtual patient scenarios in advance, the model identifies and cushions against potential harmful effects, allowing safe implementation of CGM-based therapies.
3Productivity
If simple glucose-insulin models are used, then computational requirements are lower, but the ability to describe patient physiology and variability is insufficient
Solution Approach 1:
The glucose-insulin model uses dynamic equations that continuously adjust glucose and insulin concentrations based on meal intake, physical activity, and insulin administration. The model incorporates time-varying parameters to capture the dynamic nature of patient physiology rather than using static averages.
Solution Approach 2:
The model changes key physiological parameters individually for each virtual patient based on their specific characteristics (age, weight, insulin sensitivity, carbohydrate ratio). This parameter customization allows the model to accurately describe diverse patient populations while maintaining computational efficiency through standardized mathematical frameworks.
4Ease of manufacture
If population model parameters are used, then the model is easier to implement, but it cannot account for inter- and intra-day patient variability
Solution Approach 1:
The model transitions from using uniform population parameters to assigning locally customized parameters to each virtual patient. Each patient receives individualized values for insulin sensitivity, carbohydrate-to-insulin ratio, and basal insulin rate, allowing the model to capture inter-patient variability while maintaining the same overall model structure.
Solution Approach 2:
The model pre-calculates and stores individualized patient parameters before running simulations. By preparing customized parameter sets in advance for each virtual patient scenario, the model can quickly simulate multiple days of diabetes management for diverse patient populations without requiring complex real-time calculations.
Data Source
AI summary
A mathematical model of type 1 diabetes (T1D) patient decision-making can be used to simulate, in silico, realistic glucose/insulin dynamics, for several days, in a variety of subjects who take therapeutic actions (e.g. insulin dosing) driven by either self-monitoring blood glucose (SMBG) or continuous glucose monitoring (CGM). The decision-making (DM) model can simulate real-life situations and everyday patient behaviors. Accurate submodels of SMBG and CGM measurement errors are incorporated in the comprehensive DM model. The DM model accounts for common errors the patients are used to doing in their diabetes management, such as miscalculations of meal carbohydrate content, early/delayed insulin administrations and missed insulin boluses. The DM model can be used to assess in silico if/when CGM can safely substitute SMBG in T1D management, to develop and test guidelines for CGM driven insulin dosing, to optimize and individualize off-line insulin therapies and to develop and test decision support systems.