<a><b>Background</b>: We developed a meal
detection algorithm for the artificial pancreas that detects unannounced meals
and delivers an automatic insulin bolus.</a>
<p> </p>
<p><b>Methods</b>: We conducted a randomized
crossover trial in 11 adolescents aged 12-18 years with HbA1c≥7.5% who missed≥1
bolus in the past six months. We compared (i) CSII, (ii) artificial pancreas
(AP), and (iii) artificial pancreas with a meal detection algorithm (AP+MDA).
Participants underwent three 9-hour interventions involving breakfast with a
bolus and lunch without a bolus.</p>
<p> </p>
<p><b>Results</b>: In AP+MDA, the meal detection time
was 40.0 [40.0–57.5] minutes. Compared to CSII, AP+MDA decreased the 4-hour
post-lunch iAUC from 24.1±9.5 h.mmol/L
to 15.4±8.0 h.mmol/L
(p=0.03). iAUC did not differ between AP+MDA and AP (19.6±10.4 h.mmol/L, p=0.21) nor between AP and CSII (p=0.33). The AP+MDA
reduced time>10mmol/L (58.0±26.6%) compared to CSII (79.6±27.5%, p=0.02) and AP (74.2±20.6%, p=0.047). </p>
<p> </p>
<p><b>Conclusions</b>: The AP+MDA improved glucose control after an unannounced meal.</p>