Background: Closed-loop (CL) systems direct insulin delivery based on continuous glucose monitor (CGM) sensor values. CGM accuracy varies with sensor life, being least accurate on day 1 of sensor insertion. We evaluated the effect of sensor life (enhanced Enlite, Medtronic MiniMed, Northridge, CA) on overnight CL.
Methods: In an open-label, randomized, 2-period, inpatient crossover pilot study, 12 adolescents on insulin pump (age 16.7 ± 1.9 years; HbA1c 66 ± 10 mmol/mol) attended a clinical research facility on 2 overnight occasions. In random order, participants received CL on day 1 or on day 3-4 after sensor insertion. During both periods, glucose was automatically controlled by a model predictive control algorithm informed by sensor glucose. Plasma glucose was measured every 30 to 60 min.
Results: During overnight CL (22:30 to 07:30), the proportion of time with plasma glucose readings in the target range (3.9-8.0 mmol/l, primary endpoint) when initiated on day 1 of sensor insertion vs day 3-4 were comparable (58 ± 32% day 1 vs 56 ± 36% day 3-4; P = .34), and there were no significant differences between interventions in terms of mean plasma glucose ( P = .26), percentage time above 8.0 mmol/l ( P = .49), and time spent below 3.9 mmol/l ( P = .93). Sensor accuracy varied with sensor life (mean absolute relative difference 19.8 ± 15.0% on day 1 and 13.7 ± 10.2% on day 3 to 4). Sensor glucose tended to under-read plasma glucose inflating benefits of CL on glucose control.
Conclusions: In spite of differences in sensor accuracy, overnight CL glucose control informed by sensor glucose on day 1 or day 3-4 after sensor insertion was comparable. The model predictive controller appears to mitigate against sensor inaccuracies.
Keywords: artificial pancreas; closed loop; continuous glucose monitoring; sensor accuracy; sensor life.