A Meal Detection Algorithm for the Artificial Pancreas: A Randomized Controlled Clinical Trial in Adolescents With Type 1 Diabetes.

Abstract

We developed a meal detection algorithm for the artificial pancreas (AP+MDA) that detects unannounced meals and delivers automatic insulin boluses. We conducted a randomized crossover trial in 11 adolescents aged 12-18 years with HbA_1c ≥7.5% who missed one or more boluses in the past 6 months. We compared <i>1</i>) continuous subcutaneous insulin infusion (CSII), <i>2</i>) artificial pancreas (AP), and <i>3</i>) AP+MDA. Participants underwent three 9-h interventions involving breakfast with a bolus and lunch without a bolus. In AP+MDA, the meal detection time was 40.0 (interquartile range 40.0-57.5) min. Compared with CSII, AP+MDA decreased the 4-h postlunch incremental area under the curve (iAUC) from 24.1 ± 9.5 to 15.4 ± 8.0 h ⋅ mmol/L (<i>P</i> = 0.03). iAUC did not differ between AP+MDA and AP (19.6 ± 10.4 h ⋅ mmol/L, <i>P</i> = 0.21) or between AP and CSII (<i>P</i> = 0.33). The AP+MDA reduced time >10 mmol/L (58.0 ± 26.6%) compared with CSII (79.6 ± 27.5%, <i>P</i> = 0.02) and AP (74.2 ± 20.6%, <i>P</i> = 0.047). The AP+MDA improved glucose control after an unannounced meal.