Background and Aims: Though many studies report that macronutrients such as lipids and protein significantly impact the absorption rate of carbohydrates and insulin requirement, there has not been wide adoption of simulators that take macronutrients into account. This may likely be due to lack of access to macronutrient records. This study evaluated the introduction of additional macronutrient components to a carbohydrate-only simulator.

Methods: Data uploaded between 05/15/2018-10/15/2018 by 114 individuals aged 38.1±14.2 years who used the Guardian Connect continuous glucose monitoring system with the Sugar.IQ application were analyzed. Information on 10 major macronutrients was collected on 4213 meals. Models fitting using 1) Only carb as input (“Carb-only”) and 2) Carbohydrate, lipids and protein as inputs (“Multi-macro”) were conducted and compared using the Mean Average Percent Error (MAPE) and Error Grid Analysis (EGA).

Results: The “Multi-macro” model predicts sensor glucose (SG) outcome 15% more accurately than the “Carb-only” model (MAPE: 11.3% vs. 13.5%, p<0.05). The “Multi-macro” model also improved points within the Clarke A region by 6.1% (EGA: 81.6% vs. 75.5%, p<0.05).

Conclusion: Lipids and protein are macronutrients that should be considered when creating glycemic-insulin dynamic models, as both appear to significantly impact the absorption of carbohydrates and improve SG prediction accuracy surrounding meal intake.


B. Jiang: Employee; Self; Medtronic. Y. Zhong: None. P. Agrawal: Employee; Self; Medtronic MiniMed, Inc. Stock/Shareholder; Self; Medtronic. T.L. Cordero: Employee; Self; Medtronic. R. Vigersky: Employee; Self; Medtronic MiniMed, Inc.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at