Introduction: Insulin pump is nowadays one of the trendiest treatment for type 1 diabetes and is about to become one artificial pancreas (AP) major component.
Insulin pump accuracy and performances are essentials and have not been so much studied so far, especially for smallest basal rate used for pediatric treatment and gold-standard assessment method also appears limited.
Method: A leading edge assessment method based on a double measurement - direct mass flow meter and a time-stamped micro-gravimetric bench test - combined with a Kalman bayesian-based mathematical filter has been elaborated here and is proven to measure insulin pump delivering accuracy at the lowest limit of flow rate with a much better accuracy compared with previous methods.
Its advantages and performances are illustrated while assessing the delivery precision of an already-marketed insulin pump at several flow rate. Mean absolute relative dispersion (MARD) is used to indicate overall errors for each 15-minutes error dose of a 8-hours experiments.
Results: The new proposed methodology offers a double reading - volume and flow rate - within a much more timewise precise measurement (0.03Hz to 10Hz): direct instantaneous basal rate is displayed for the first time. Measurement accuracy is also considerably improved: pediatric basal rate is read with a 1.2%error instead of 15.3% error before. Accuracy is shown to be different regarding the flowrate value:Pump#1(2UI/h): MARD = 12.7%, Pump#1(0.5UI/h): MARD = 20.4%Pump#1(0.1UI/h): MARD = 65.7%
Conclusion: This innovative method to assess insulin pump administration highlights an imprecision of insulin delivery, especially for lowest basal rate as used in pediatric. Error identification might increase clinical impact and feed AP control algorithm while anticipating delivering errors.
S. Girardot: Employee; Self; Air Liquide. F. Mousin: Employee; Self; Air Liquide. S. Hardy: Employee; Self; Air Liquide. J. Riveline: Board Member; Self; Abbott, Lilly Diabetes, Novo Nordisk Inc. Consultant; Self; Novo Nordisk Inc. Research Support; Self; Air Liquide, Amgen Inc. Speaker's Bureau; Self; Abbott, Lilly Diabetes, Sanofi.