Page 327 - Glucose Monitoring Devices

334 CHAPTER 16 The dawn of automated insulin delivery

Speeding up the process: the creation of an FDA approved
simulator

Evaluation of hybrid closed-loop systems in animal models before human studies
take an exceptionally long time and be very costly. Kovatchev and colleagues
were able to show that in silico preclinical studies could be performed and replace
animal trials through the University of Virginia-University of Padova type 1 diabetes
simulator [37]. Based on these studies, the FDA approved the use of in silico studies
to test closed-loop studies before clinical trials in humans. After this achievement,
in silico studies projected to take years with animal trials were able to be tested
in less than 6 months [38].

Early studies aimed at closing the loop
The first feasibility study of subcutaneous closed-loop automated insulin delivery
was reported by Steil in 2006 and used a PID algorithm with a glucose setpoint
of 120 mg/dL (6.7 mmol/L) [39]. This 30-h inpatient study compared data from
open-loop (OL) therapy to what was attained with this fully closed-loop (FCL)
system [39]. Participants increased time in the range of 70e180 mg/dL
(3.9e10 mmol/L) from 63% in OL to 75% in FCL, with no episodes of severe
hypoglycemia and similar rates of hypoglycemia in both groups [39].
Given the delay in time to peak insulin action with rapid-acting insulin analogs
and the tendency for postprandial hyperglycemia noted with the FCL, Weinzimer
and colleagues evaluated the feasibility of using a hybrid closed-loop approach using
a PID algorithm in adolescents [40]. Seventeen adolescents with T1D underwent an
initial OL assessment of glycemic control with a blinded sensor (Medtronic Datalog-
ger CGM) with eight undergoing a 34-h FCL study. The remainder of the partici-
pants received 25%e50% of their usual mealtime bolus 15-min before eating
[40]. This latter group was termed “hybrid” closed-loop, as there was a requirement
to bolus for the meals [40]. Although overall mean glucose levels were similar
between the two groups, postprandial hyperglycemia was more pronounced in the
FCL group when compared to the hybrid closed-loop group (226 51 mg/dL vs.
194 47 mg/dL, respectively) [40]. Based on results from this trial, many have
since adopted a hybrid closed-loop approach for other systems in development.

Control in clinic: the first closed-loop studies in rigorous
research environments
To first evaluate the safety and feasibility of closed-loop insulin delivery, studies
were conducted in very controlled research environments. This afforded investiga-
tors the opportunity to use bedside plasma glucose values as a means to verify sensor
accuracy while also allowing close scrutiny of how plasma glucose levels were being