Since the approval of continuous glucose monitoring (CGM) devices for insulin dosing in type 1 diabetes (T1D) treatment, one open issue is how to exploit CGM-derived information, e.g. blood glucose (BG) rate-of-change (ROC), to improve the calculation of meal-time insulin bolus provided by the standard formula (SF). The aim of this work is performing an in-silico comprehensive comparison of literature methods proposed so far for such a purpose.
The UVa/Padova T1D Simulator was used to generate data of 100 virtual patients undergoing multiple single meal scenarios with different preprandial BG (70, 120, 180 mg/dL) and ROC (-2, -1, 1, 2 mg/dL/min). We considered six literature methodologies that adjust SF according to future CGM predictions (Scheiner, Pettus/Edelman), by a percentage modulation (Buckingham), adding/subtracting a fixed insulin quantity (Klonoff), also based on insulin sensitivity (Aleppo/Laffel, Ziegler). The quality of glycemic control has been quantified as the difference between the blood glucose risk indexes (ΔBGRI) of final glucose profiles (SF vs literature methods).
Results are reported in Figure 1.
Literature methods accounting for ROC are effective, but no method prevails over the others. Particularly, the improvement is limited to the combination low BG/positive ROC, and high BG/negative ROC. As such, further investigations are required to optimize insulin therapy using CGM information.