Understanding glycemic control in everyday life: The importance of accurate quantification of postprandial response profiles for different types of food

Objective Clinicians commonly use tests such as clamps and oral/intravenous glucose tolerance tests (OGTT/IVGTT) to assess the glycemic status of diabetes patients. These rather artificial situations, however, provide very limited insight into the dynamics of postprandial glucose profiles in everyday life. Mathematical models developed to better understand the glucose-insulin system are typically calibrated on data of OGTTs or standardized meals. Those models can therefore not be used to provide food-specific predictions. The only available measure of postprandial glucose profiles is the Glycemic Index (GI). Although the GI reflects the area under the 2h-postprandial glucose curve, it does not provide information on the height and duration of the postprandial glucose peak, which are highly important factors in glycemic control. Therefore, the GI does also not provide enough information to better understand glycemic control in everyday life. We propose an innovative approach to resemble and understand physiological situations more accurately by integrating postprandial meal data in the mathematical model of the Eindhoven Diabetes Education Simulator (E-DES). This tool will serve as a personalized test environment for diabetes patients, helping them to gain insight into the effect of everyday activities on daylong glucose and insulin profiles. Methods The E-DES model consists of four coupled integral-differential equations describing the glucose and insulin balance in three compartments (blood, gut and interstitial fluid). To adjust the model to incorporate different types of food, a large scale data integration approach was employed. The entire model was first calibrated on OGTT data in normal glucose tolerant subjects. Next, the model predictions were adjusted by re-estimating the model parameters that govern the digestion and absorption related processes on postprandial response data of individual food products and composite meals. Furthermore, the robustness of the calibrated model was checked through a sensitivity analysis of the model parameters. Results Meta-analysis of postprandial literature data upon a variety of consumed food revealed highly fluctuating kinetic profiles of postprandial glucose and insulin responses. Peak shapes ranged from sharp (fast absorbed carbohydrates) to gradual and of longer duration (slowly absorbed carbohydrates). By re-estimating only the four food-related parameters, our model successfully reproduced postprandial glucose and insulin profiles for 35 different food types, ranging from single food products like plain bread to large composite meals including breakfast, lunch, snacks and dinner. Conclusion This study showed that our model could successfully be applied to describe specific postprandial responses to a variety of food intake. This novel approach minimizes the gap between application of mathematical models in artificial (experimental) situations and in normal physiological conditions. Furthermore, the inclusion of food-specific postprandial glucose predictions in everyday life conditions is an essential component in the development of the E-DES