Incorporating different food products and composite meals in the Eindhoven Diabetes Education Simulator
Anne Maas,1,2 Yvonne Rozendaal,2 Carola van Pul,1 Ward Cottaar,3 Peter Hilbers,2 Natal van Riel,2 Harm Haak.1
1Department of Internal Medicine, Máxima Medical Center, Eindhoven. 2Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven. 3Design and Technology of Instrumentation, Stan Ackermans Institute, Eindhoven. Email: .
Background: Diabetes education is mainly based on one-on-one patient-health care provider contact. This is costly, time-consuming and gives the patient no room for practice. We want to address these issues by creating the Eindhoven Diabetes Education Simulator, which uses a physiology-based mathematical model to predict glucose and insulin concentrations for patients with diabetes type 1 and 2 over a 2-4 hour time period after intake of food and/or insulin. In our current model food is entered in the form of carbohydrate content. The goal of this study was to incorporate different food products and composite meals for healthy persons, since different food types will elicit different glucose responses.
Methods: A literature search was performed for datasets of different food products using (combinations of) the following search terms: healthy, mixed meal, glucose, insulin, glycemic response, glycemic index, and looking for cross-references. We included any dataset for which glucose ánd insulin concentrations were measured on at least 5 time points after food ingestion in healthy subjects. Healthy was defined as normal glucose tolerant, normal insulin sensitive, normotensive, normal HbA1c, non-obese (BMI< 30 kg/m2), no family history of diabetes, not pregnant, and free of apparent diseases and medication. Our model was fitted to the different datasets using a non-linear least squares algorithm.
Results: We have fitted our model to 57 separate datasets (from 18 publications including 220 subjects, references available on request). For 35 of these datasets we obtained a model fit that described the dataset well, of which five are shown in Figure 1. In the cases that we could not obtain a good fit, there usually were a limited number of data points available.
Conclusion: The Eindhoven Diabetes Education Simulator is able to simulate postprandial glucose and insulin concentrations for healthy persons for 35 different food products and composite meals.
Figure 1: Glucose (left) and insulin (right) concentrations after intake of five different meals. The crosses with error bars are the data and standard deviation obtained from literature, the solid line shows the corresponding model fit.
Figure 2: Glucose (left) and insulin (right) concentrations after intake of five different meals. The crosses with error bars are the data and standard deviation obtained from literature, the solid line shows the corresponding model fit.