International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

journal homepage: www.elsevier.com/locate/ijpharm

Glucose-responsive insulin delivery for type 1 diabetes: The artificial

pancreas story
⁎
Lia Ballya,b, , Hood Thabitc,d, Roman Hovorkae,f
a
Department of Diabetes, Endocrinology Clinical Nutrition & Metabolism, Inselspital, Bern University Hospital, University of Bern, Switzerland
b
Department of General Internal Medicine, Inselspital, Bern University Hospital, University of Bern, Switzerland
c
Central Manchester University Hospitals NHS Foundation Trust, Manchester Academic Health Science Centre, Manchester, United Kingdom
d
Division of Diabetes, Endocrinology and Gastroenterology, Faculty of Biology, Medicine and Health, University of Manchester, United Kingdom
e
Wellcome Trust-MRC Institute of Metabolic Science, University of Cambridge, Cambridge, United Kingdom
f
Department of Paediatrics, University of Cambridge, Cambridge, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: Insulin replacement therapy is integral to the management of type 1 diabetes, which is characterised by absolute
Insulin pump therapy insulin deficiency. Optimal glycaemic control, as assessed by glycated haemoglobin, and avoidance of hyper- and
Continuous glucose monitoring hypoglycaemic excursions have been shown to prevent diabetes-related complications. Insulin pump use has
Closed-loop increased considerably over the past decade with beneficial effects on glycaemic control, quality of life and
Artificial pancreas
treatment satisfaction. The advent and progress of ambulatory glucose sensor technology has enabled continuous
Control algorithm
glucose monitoring based on real-time glucose levels to be integrated with insulin therapy. Low glucose and
predictive low glucose suspend systems are currently used in clinical practice to mitigate against hypoglycaemia,
and provide the first step towards feedback glucose control. The more advanced technology approach, an ar-
tificial pancreas or a closed-loop system, gradually increases and decreases insulin delivery in a glucose-re-
sponsive fashion to mitigate against hyper- and hypoglycaemia. Randomised outpatient clinical trials over the
past 5 years have demonstrated the feasibility, safety and efficacy of the approach, and the recent FDA approval
of the first single hormone closed-loop system establishes a new standard of care for people with type 1 diabetes.

1. Introduction insulin analogues into clinical practice has allowed modern insulin
therapy to become more flexible to suit individuals’ lifestyle. Still, the
Type 1 diabetes represents 5–10% of diabetes cases worldwide. The lack of a physiological-like feedback for insulin delivery prevents re-
prevalence over the past 20 years however has doubled, increasing the commended therapy targets to be achieved as daily insulin require-
burden of the condition for individuals and society (Atkinson et al., ments can vary widely due to complex interactions between the nutri-
2014). The cause of type 1 diabetes is an absolute insulin deficiency due tional intake, physical activity and hormonal milieu (Ruan et al., 2016).
to autoimmune destruction of beta-cells (Atkinson et al., 2014). Thus, Advances in diabetes technology over the past decades have led to
the mainstay of type 1 diabetes treatment is insulin replacement the development of continuous glucose monitoring (CGM) which
therapy with the goal of achieving near normoglycaemia to reduce the measure the real-time interstitial glucose concentration (Rodbard,
risk of diabetes-related micro- and macrovascular complications 2016). Conventional CGM sensors are externally worn. Recently how-
(Nathan et al., 1993). However, tight glycaemic control increases the ever, an implantable CGM with sensor life of up to six months has been
risk of hypoglycaemia with potentially severe acute and long-term se- developed [http://www.senseonics.com/investor-relations/news-
quelae (Frier, 2014). releases/2017/09-12-2017-213230489]. Coupling of CGM to insulin
Successful management of type 1 diabetes entails frequent glucose delivery via a control algorithm, known as a closed-loop system or the
monitoring and insulin dose adjustment. Conventional insulin therapy artificial pancreas, has enabled device-centred automated glucose-re-
management is based on the administration of prandial insulin at meal- sponsive insulin delivery to be progressed (Thabit and Hovorka, 2016).
times, and basal insulin to control glycaemia during fasting and be- This allows modulation of insulin-delivery to address the constantly
tween meals (Fig. 1). The introduction of insulin pump therapy and changing insulin requirements and to reduce the burden of daily self-

⁎
Corresponding author at: Department of Diabetes, Endocrinology, Clinical Nutrition and Metabolism and Department of General Internal Medicine, Inselspital, Bern University
Hospital, University of Bern, Freiburgstrasse, 3010 Bern, Switzerland.
E-mail address: lia.bally@insel.ch (L. Bally).

https://doi.org/10.1016/j.ijpharm.2017.12.022
Received 4 September 2017; Received in revised form 4 December 2017; Accepted 10 December 2017
0378-5173/ © 2017 Elsevier B.V. All rights reserved.

Please cite this article as: Bally, L., International Journal of Pharmaceutics (2017), https://doi.org/10.1016/j.ijpharm.2017.12.022
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

Fig. 1. The concept of basal bolus insulin therapy. Basal insulin provides
background insulin for periods between meals and overnight. Bolus insulin
targets meal-related glucose excursions.

management. pumps have an integrated short infusion cannula and are attached di-
In this review article, we outline the present landscape of insulin rectly to the body with an adhesive patch. However, the remote con-
pump therapy and introduce the concept of glucose-responsive auto- troller might get lost or forgotten disabling bolus application.
mated insulin delivery or closed-loop systems. Current limitations and The use of insulin pumps can be augmented by the use of CGM. The
future developments of closed-loop therapy are also discussed. approach is known as sensor-augmented pump therapy (SAP).
Interstitial glucose is measured every 1–5 min, and glucose values are
2. Insulin pump therapy either sent directly to the pump, or to separate hand-held receiver. User
input is needed to manually adjust insulin pump delivery in conven-
Continuous subcutaneous insulin infusion, commonly known as tional SAP systems based on CGM or capillary glucose measurements.
“insulin pump therapy”, was introduced in the 1970s (Pickup et al.,
1978). It is increasingly used in clinical practice as an alternative to 2.2. Clinical evidence in support of insulin pump therapy
multiple daily injection (MDI) therapy. The increased uptake of insulin
pump therapy is in part due to the Diabetes Control and Complications A meta-analysis of relatively short-term randomised controlled trials
Trial (DCCT, 1993), which showed significant decrease in micro- showed a mean reduction in glycated haemoglobin (HbA1c) of
vascular disease risk with intensive insulin therapy (Nathan et al., 0.3–0.6% (3–6 mmol/mol) (Pickup and Sutton, 2008; Misso et al.,
1993). The evidence from clinical studies evaluating the safety, clinical 2010) during insulin pump therapy compared to MDI. This was ac-
and cost effectiveness of insulin pump therapy has been published companied by a 20% reduction in total daily insulin dose. Pooled data
(Pozzilli et al., 2015). Guidelines and recommendation for insulin pump from randomised controlled and observational studies, in which users
therapy are available for health care professionals involved in the switched from MDI to insulin pump therapy, showed that the greatest
management of people with type 1 diabetes (Peters et al., 2016). improvement in HbA1c occurred in those with the highest baseline
HbA1c (Pickup and Sutton, 2008); those with baseline HbA1c of 9%
2.1. Insulin pump features and functionalities (75 mmol/mol) on MDI showed an HbA1c reduction of approximately
1.5% (16 mmol/mol) when switching to insulin pump therapy whilst
The advantage of insulin pumps over injection therapy is the ability the reduction was less than 0.5% (5 mmol/mol) in those with baseline
to set different pre-programmed basal rates of insulin delivery HbA1c of 7.5% (58 mmol/mol).
throughout the day to match the individual’s varying diurnal insulin A further reduction in HbA1c was observed when conventional in-
needs (Pickup, 2012). Moreover, insulin pumps allow for meal insulin sulin pump users were switched to SAP with the degree of HbA1c im-
to be delivered using different bolus profiles, to suit meals of different provement again dependent on baseline HbA1c (Pickup et al., 2011).
macronutrient composition and absorption; insulin boluses can either Regular glucose sensor use was also a determinant factor. An observa-
be delivered immediately (known as standard, immediate or step tional study related to long-term outcomes on insulin pump therapy
bolus), or over several hours (square or extended wave bolus), or as a demonstrated sustainability of glycaemic control over prolonged use of
combination of the two (dual wave bolus). Modern insulin pumps have insulin pump therapy (Beato-Víbora et al., 2015).
integrated bolus calculators used by the user to determine the appro- The majority of clinical studies have shown that the burden and risk
priate insulin bolus dose, by inputting pre-meal blood glucose levels of hypoglycaemia are ameliorated by insulin pump therapy compared
and estimated carbohydrate intake. The bolus calculator provides the to MDI. A meta-analysis of studies evaluating those at increased hy-
recommended bolus dose by taking into account the pre-programmed poglycaemia risk showed overall a 75% reduction in severe hypogly-
target glucose, insulin sensitivity, insulin to carbohydrate ratio and caemia events requiring third party assistance, when users were swit-
amount of active insulin remaining (known as insulin on board) (Zisser ched from MDI to insulin pump therapy (Pickup and Sutton, 2008), and
et al., 2008). Most insulin pump devices have alarms to alert the user in the relative risk reduction was at least 10-fold in those with the highest
case of an occlusion or other technical malfunctions to increase safety baseline risk. Hypoglycaemia reduction was observed to a lesser extent
of the approach. Glucose, insulin and user specific data, such as bo- in children, likely due to the shorter duration of diabetes and a lower
lusing behaviour, can be downloaded and used by users and healthcare baseline frequency of severe hypoglycaemia (Pickup and Sutton, 2008).
professionals to identify areas requiring further attention to optimise Hypoglycaemia occurrence has been shown to be related to increased
glycaemic control. glucose variability (Kilpatrick et al., 2007). Insulin pump therapy has
Conventional insulin pump devices deliver insulin via plastic been shown to reduce within-day and between-day glycaemic varia-
tubing, and teflon or steel cannula changed every 2–3 days and inserted bility, which may in part contribute towards ameliorating the risk of
subcutaneously. Disadvantages include the aesthetic concerns and the hypoglycaemia.
risk of kinking or disconnection of the tubing. Tubingless or patch A notable difference between older studies such as the Diabetes

2
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

Control and Complications Trial and more recent studies which com- events/100 patient years, p = .013). The number of participant drop-
pared MDI with insulin pump therapy is the wider use of insulin ana- outs however was relatively high (24/30 and 12/30 in the DiaPort and
logues during the latter. As such, reported outcomes from these studies subcutaneous insulin pump group, respectively), which the authors
and meta-analysis may not be directly comparable and need to be in- attributed to participants’ fear of potential complications of in-
terpreted with caution (Pickup and Sutton, 2008; Misso et al., 2010; traperitoneal delivery, and reluctance to use it. The commonest adverse
Weissberg-Benchell et al., 2003). In addition, those with hypogly- event is infection around the percutaneous port. The second generation
caemia-unawareness were often excluded and therefore findings may DiaPort was introduced in 2011, re-designed with a Dacron cuff around
not be generalizable to those with significant burden of hypoglycaemia the percutaneous access port to provide a barrier to infection and better
(Pickup, 2012). The incremental benefits of insulin pump therapy over stabilization of the port at the implantation site, which could poten-
MDI using novel long-acting insulin analogues which have shown less tially minimise the incidence of serious or persistent infections. Non-
pharmacodynamics variability, in conjunction with state-of-the-art randomised single-arm studies with relatively small number of parti-
functional insulin therapy education programmes, remain unclear. cipants have been performed (i.e. ClinicalTrials.gov NCT01483352);
Improvements in the quality of life and treatment satisfaction amongst results are yet to be published.
insulin pump users however have been consistently shown (DeVries
et al., 2002). A recent study evaluated the glycaemic benefits of insulin 2.4. Limitations of insulin pump therapy
pump therapy in adults with type 1 diabetes on MDI regime already
using CGM. Participants used the G4 sensor (Dexcom, San Diego, CA, Modern insulin pumps are provided with alarms for catheter oc-
USA) and the Insulet OmniPod insulin pump (Insulet, Billerica, MA, clusion, pump mechanism failures and low battery. However, if not
USA) during the intervention period. Efficacy endpoints were assessed vigilant or well-trained, pump users could be exposed to clinically
over a period of 28 weeks and showed a mean reduction in sensor significant adverse events due to specific limitations of insulin pump
glucose by 0.8 mmo/l (14 mg/dl), and a 6% or 83 min per day increase therapy. In contrast to basal bolus MDI regime, the lack of long-acting
in time spent with sensor glucose in target range (3.9–10.0 mmol/l insulin puts insulin pump users at an increased risk of diabetic ketoa-
[70–180 mg/dl]), when insulin pump-CGM users were compared to cidosis due to insulin deficiency in the event of prolonged catheter
MDI-CGM. This came at the expense of increased time spent hypogly- displacement, occlusion or kinking. In a cross-sectional survey of pump
caemic (Beck et al., 2017). HbA1c levels remained unchanged following users, nearly half reported some form of insulin delivery problems, ei-
transition from MDI to insulin pump therapy. ther due to infusion set or pump device issues (Pickup et al., 2014). The
use of the infusion set longer than 3 days was also associated with in-
2.3. Continuous intraperitoneal insulin delivery sulin delivery and absorption issues. A comparison between steel and
teflon catheters found a 15% rate of initial infusion set failure with the
Continuous intraperitoneal insulin infusion (CIPII) may confer some latter, due to the catheters kinking on insertion (Patel et al., 2014). This
advantages due to the more favourable pharmacokinetic profile com- highlights the importance of user and health-care provider adherence to
pared to subcutaneous insulin delivery, with faster absorption rate and appropriate pump care and guidelines such as sick day rules and set
restoration of physiological portal-peripheral gradient (Schade et al., change intervals, to mitigate against common hazards of pump therapy.
1981). This led to the development of implantable insulin pumps which Post marketing surveillance programmes of pump therapy are needed
began in the 1970s. Implantable pumps were initially manufactured by to ensure safety standards are met and to improve the reliability of
several companies such as Infusaid Inc. (Norwood, MA, USA), MiniMed pump components by manufacturers. Concerns of potentially higher
Inc. (Sylmar, CA, USA) and Siemens-Elema (Solna, Sweden). MiniMed risk of DKA during insulin pump therapy was recently challenged by a
Inc. has since merged with Medtronic, and the Minimed 2007 remains systematic review which reported lower prevalence of DKA in adult
the only commercially implantable pump available at present. It is insulin pump users compared to MDI (Fazeli Farsani et al., 2017).
approved for clinical use in Europe, however, it is limited to Medtronic- Likewise, a recent population-based cohort study involving 30,579 in-
certified centres in Netherlands, Belgium, Sweden and France. dividuals aged between 1 and 20 years reported lower risk of ketoaci-
Despite their theoretical advantages, regulatory approval and im- dosis in pump users compared to insulin injection therapy (Karges et al.,
plementation of IPI in clinical practice have been limited, due to con- 2017). This however does not abdicate the need for continuous vigi-
cerns related to long-term safety, maintenance, costs and regular need lance and education of insulin pump users.
of specialist input (Bally et al., 2016). Reports and studies evaluating Health economic analyses have shown that the cost of insulin pump
implantable insulin pump use with surfactant-stabilized U-400 insulin therapy is notably higher than MDI. The annualised cost of pump
formation and clinical feasibility of IPI have been published, albeit in a therapy in the UK, which includes device and consumables expenses
small number of users and centres (Logtenberg et al., 2009; van Dijk over a warranty period, is approximately £1700 greater compared to
et al. 2015; Schaepelynck et al., 2011). MDI. In addition, the educational input for pump users is difficult to
The Accu-Chek Diaport system (Roche Diabetes), which involves the quantify. Overall, the cost-effectiveness was shown to be in favour of
use of percutaneous access, is a less invasive approach of CIPII (Liebl insulin pump therapy relative to MDI, due to an projected 1.2%
and Frei, 2003). It composes of a catheter which is inserted in the (2 mmol/mol) HbA1c reduction (Cummins et al., 2010). Others have
peritoneal space and fixed to a surgically implanted metallic body at- shown pump therapy to be cost-effective when health economic model
tached to the skin. The outer port (located externally approximately such as the Centre for Outcomes Research (CORE), which accounts for
2–3 mm from the skin), is connected to an external insulin pump device the quality of life in addition to HbA1c, was used (Roze et al., 2015). It
(Accu-Chek Spirit Combo pump, Roche Diagnostics, Mannheim, Ger- is important to note that these models currently do not evaluate the
many). Infused insulin through the Diaport system is a regular U-100 burden of hypoglycaemia on the individual and society, such as loss of
insulin (Insuman Infusat, Sanofi, Paris, France) as rapid-acting insulin productivity at work or school, and mental health issues associated with
analogues have not been approved for intraperitoneal delivery. In a fear of hypoglycaemia.
multi-centre cross-over design study, sixty participants with type 1
diabetes were randomised to intraperitoneal insulin infusion using 3. Insulin pump therapy with threshold suspend features
DiaPort system with regular insulin (Insuman U-100), or subcutaneous
insulin pump therapy with lispro for 12 months (Liebl et al., 2009). The SAP has been shown to lead to a significant improvement in dia-
primary endpoint, incidence of hypoglycaemia, was comparable be- betes management allowing the user to set personalised alarms.
tween both treatments groups (p = .91), whereas the number of severe However, users do not always take the appropriate actions following an
hypoglycaemia events was halved during the DiaPort use (34.8 vs. 86.1 alert. To reduce the risk of hypoglycaemia, approaches that

3
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

automatically suspend insulin delivery have been developed and tested

assistance). 9.5 vs. 34.2 events/100 patient months (incidence ratio 3.6, p < 0.001), LGS vs.
Combined incidence of severe (seizure or coma) and moderate hypoglycaemia (third party

(54% reduction, p < 0.001), 4–10year-olds: 3.1 vs. 6.2% (50% reduction, p < 0.001), PLGS
Nocturnal sensor glucose AUC < 3.6 mmol/l (65 mg/dL). 54.4 vs. 87.0 mmol/lxmin (980

% of nights with ≥ 1 sensor glucose value ≤ 3.3 mmol/l (60 mg/dl). 21% vs. 33% (OR
% of time with sensor glucose < 3.9 mmol/l (70 mg/dl). 11–14year-olds: 4.6 vs. 10.1%
in randomised clinical trials (Table 1).

Number of hypoglycaemic events < 3.6 mmol/l (65 mg/dl): 4.4 vs.7.4, PLGS vs. SAP
% of nights with sensor glucose ≤ 3.9 mmol/l (70 mg/dl). 16 vs. 30%, PLGS vs. SAP
3.1. Low glucose suspension of insulin delivery

Automatic cessation of insulin delivery at a pre-set threshold sensor

glucose value is referred to as low glucose suspend (LGS). A LGS event
initiates up to 2 h suspended insulin delivery; insulin delivery can re-
sumed earlier by the user. The Paradigm VEO system (Medtronic
vs. 1568 mg/dlxmin), LGS vs. SAP (37.5% lower, p < 0.001)

MiniMed, Inc, Northridge, CA, USA) was the first pump with built-in
LGS feature and introduced in Europe in 2009. The use of LGS over an
extended period of time in hypoglycaemia prone individuals reduced
the risk and frequency of nocturnal hypoglycaemia without increasing
HbA1c (Bergenstal et al., 2013; Ly et al., 2013). In 2013, the ASPIRE In-
Home Study Group evaluated the effect of SAP with LGS (Bergenstal
et al., 2013) compared to SAP alone, on nocturnal hypoglycaemia and
0.52, p < 0.001), PLGS vs. SAP.
Primary/co-primary outcome(s)

HbA1c levels in participants at increased risk of nocturnal hypogly-

caemia (experienced ≥2 nocturnal hypoglycaemic events during the 2-
conventional pump therapy

week run-in phase). Mean AUC for nocturnal hypoglycaemic events was
significantly decreased by 32% in the LGS group after 3 months com-
pared to control. Time spent hypoglycaemic was also significantly
lower in LGS compared to control. Others have shown that the initial
(p = 0.008).

excess cost of LGS therapy is balanced by avoidance of hypoglycaemia-

vs. SAP.

related healthcare expenditures (Ly et al., 2014; Roze et al., 2016).

3.2. Predictive low glucose suspension of insulin delivery

Overnight, 3 months

Day and night, 2

Day-and-night, 6
Study duration

Algorithms that suspend insulin delivery before a threshold low

glucose value is reached have been developed and are referred to as
21 nights

42 nights

42 nights
months

predictive low glucose suspend (PLGS) systems (i.e. Medtronic MiniMed

weeks

640G pump) aiming to reduce further the burden of hypoglycaemia.

Briefly, the 640G algorithm suspends insulin delivery when the glucose
level is predicted to drop to a level equal to or lower than a pre-set
Conventional pump

threshold within 30min. Once the predicted glucose is safely above the
threshold or after 2 h, insulin delivery is resumed (Fig. 2). Randomised
Comparator

clinical trials with PLGS have reported reduced nocturnal hypogly-

therapy

caemia burden compared to SAP. PLGS decreased the burden of hy-

LGS, Low Glucose Suspend; PLGS, Predictive Low Glucose Suspend; SAP, Sensor-Augmented Pump Therapy.
SAP

SAP

SAP

SAP

SAP

poglycaemia as measured by the median hypoglycaemia area under the

curve by 81%, and hypoglycaemia lasting > 2 h was reduced by 74% in
Randomised controlled home studies of threshold suspend (low glucose and predictive low glucose).

Intervention

a 42-night randomised trial involving adolescents and adults with type

1 diabetes (Maahs et al., 2014). A recent analysis of PLGS use in clinical
PLGS

PLGS

PLGS

PLGS

practice showed reduction of hypo- and hyperglycaemic events during

LGS

LGS

both night and daytime (Zhong et al., 2016). Table 1 summarises ran-
domised controlled trials evaluating efficacy of LGS and PLGS in type 1
95 (46 LGS, 49

100 (50 PLGS,

147 (121 LGS,

diabetes.
pump-only)
126 SAP)

50 SAP)

4. Closed-loop insulin delivery

19

45

45
N

Closed-loop systems or the artificial pancreas are more advanced

than LGS and PLGS system and autonomously and gradually increase
Children, adolescents and adults

and decrease subcutaneous insulin delivery according to real-time

Adolescents and adults with
nocturnal hypoglycaemia.

sensor glucose levels. Such a glucose-responsive insulin technology

Children and adolescents
Adolescents and adults

aims to mimic physiology of the pancreatic beta cell by providing in-

sulin delivery commensurate with acute individual metabolic needs.
Study population

Closed-loop insulin delivery has the potential to reduce the burden of

Adolescents

diabetes self-management.
Children

4.1. Closed-loop system components

Bergenstal et al. (2013)

Battelino et al., (2017)

4.1.1. Overview
Maahs et al., (2014
Buckingham et al.,

Closed-loop systems consist of a glucose sensor, an insulin pump and

Buckingham et al.
Ly et al., (2013)

a control algorithm which is either embedded in the pump or residing

in a separate device, e.g. a smartphone (Fig. 3). These components
(2013)

(2015)
Reference

communicate wirelessly. The key component of a closed-loop system is

Table 1

the control algorithm, which directs insulin delivery according to real-

time glucose levels while accounting for inter- and intra-subject

4
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

Fig. 2. Two predictive low glucose suspend events (including alarms) are shown within the red square. The upper panel shows the sensor glucose profile. Insulin delivery profile is shown
in the lower panel.

variability, inherent sensor and insulin delivery errors as well as kinetic operate autonomously without the need for user-initiated prandial in-
delays. Various algorithms have been developed and the two most sulin boluses. The performance of the latter is limited by the delayed
commonly used are the proportional-integral-derivative control and the action profile of subcutaneous insulin delivery (Weinzimer et al., 2008)
model-predictive control approach (Doyle et al., 2014). and is currently not pursued commercially. The hybrid closed-loop
system overcomes delays through bolus delivery and announcement to
4.1.2. Proportional-integral-derivative control the control algorithm of meals (Fig. 4). The hybrid systems are the focus
The proportional-integral-derivative algorithms adjust insulin de- of current translational research.
livery by considering deviations from a target glucose level (propor- Closed-loop systems using either single-hormone (insulin only) or
tional component), the area under the curve between the measured and dual-hormone (insulin and glucagon) have been studied.
target glucose level (integral component), and the rate of change in the
measured glucose levels (derivative component). 4.2.2. Co-delivery of glucagon
Glucagon counteracts insulin action through mobilization of the
4.1.3. Model predictive control endogenous glucose production from the liver (Haidar et al., 2016a).
The model predictive control algorithms employ a mathematical The aim of glucagon delivery in closed-loop systems is either to alle-
model of glucose regulation to predict glucose excursions. Insulin de- viate solely the risk of hypoglycaemia, or to enable insulin to be de-
livery is calculated by minimizing the difference between model-pre- livered more aggressively and counteract insulin overdelivery (Taleb
dicted glucose-concentrations and the target glucose levels over a pre- et al., 2017). The control algorithm of the former is similar to that of a
specified prediction time horizon. single hormone approach. Glucagon is given intermittently as mini
Other clinically evaluated control approaches include the fuzzy boluses by a separate pump chamber, with daily doses ranging up to
logic approach which modulates insulin delivery on the basis of ap- 1 mg/day (Taleb et al., 2017). Currently available glucagon formula-
proximate rules to express the empirical knowledge of diabetes prac- tions, such as GlucaGen® (Novo Nordisk, Bagsvaerd, Denmark), consist
titioners (Nimri et al., 2014aa). Most control algorithms include safety of lyophilized powder that requires daily reconstitution due to protein
modules to constrain insulin delivery, limiting the amount of insulin on aggregation and degradation (Caputo et al., 2013), and as such addition
board or the maximum rate of insulin delivery, or suspending insulin of glucagon to a closed-loop system could therefore impose further
delivery when glucose levels are low or decreasing. burden and complexity to the user.

4.2. Types of closed-loop systems 4.2.3. Co-delivery of incretin-based therapies

Apart from glucagon, use of other hormone-based therapies in
4.2.1. Overview combination with closed-loop insulin delivery have been proposed. The
The two main approaches are (i) hybrid closed-loop systems that adjunctive use of amylin analogues and incretin-based therapies may
require for optimal performance manual input through user-initiated attenuate peak meal-related glucose appearance and improve post-
insulin boluses at meal times and (ii) fully closed-loop systems which prandial glycaemia by delaying gastric emptying and suppressing meal-

Fig. 3. Panel A: A closed-loop system comprising a

glucose sensor (black rectangle on the left-hand side
of the abdomen), an insulin pump (device in the
pocket), and a mobile-sized device containing the
control algorithm (in user’s hand). Each component
communicates with the other wirelessly (adapted
from Hovorka, 2011). Panel B: The closed-loop
system mimics the physiological feedback normally
provided by the pancreatic β-cell.

5
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

Fig. 4. Glucose levels and insulin delivery profiles

during a hybrid single hormone closed-loop study
(Bally et al., 2017) The bold red line shows the
continuous subcutaneous glucose trace. Control al-
gorithm-directed insulin infusion rates during
closed-loop are denoted by the thick blue line. Ver-
tical arrows indicate meals and snacks (orange ar-
rows) and blue vertical lines illustrate insulin boluses
administered by the user at meal times. Horizontal
dashed red lines indicate the target glucose range
from 3.9 to 10 mmol/l. (For interpretation of the
references to colour in this figure legend, the reader
is referred to the web version of this article.)

induced glucagon response (Hinshaw et al., 2013). system are both undergoing pilot clinical evaluations (NCT02849288
Pramlintide is a synthetic analogue of amylin and has been shown to and NCT02897557, respectively). Closed-loop systems developed by
improve postprandial-prandial glucose control when injected pre-meal Type Zero Inc (http://typezero.com/) and the DIABELOOP consortium
during single-hormone closed-loop operation (Weinzimer et al., 2012). are undergoing multicentre pivotal testing (NCT02985866 and
Combining single-hormone closed-loop insulin-delivery with a GLP-1 NCT02987556, respectively). Commercialisation of a bi-hormonal
receptor agonist (liraglutide) was shown to mitigate against meal-in- closed-loop system in Europe is planned (Trevitt et al., 2016). An in-
duced glycaemic excursions whilst having an insulin-sparing effect, tegrated dual-hormone system (ILet) is currently being tested in a fea-
when compared to adjunct use of amylin (Sherr et al., 2016). The ad- sibility trial in the US (NCT02701257). Do-it-yourself closed-loop sys-
dition of DPP-4 inhibitor to single-hormone closed-loop improved tems are being developed and used by the sizable open-source
postprandial glycaemic control, compared to closed-loop alone community (https://openaps.org).
(Underland et al., 2017). Despite its benefit however, concerns re-
garding gastrointestinal side-effects remain.
4.4. Clinical evidence

4.3. From prototypes to commercial hybrid closed-loop pump Over the past decade, clinical studies have progressed from con-
trolled research facility (Hovorka et al., 2011) and supervised transi-
Several closed-loop prototypes have been developed by different tional settings (Del Favero et al., 2015; Phillip et al., 2013) to free-living
academic and commercial entities. Academic prototypes host the con- home studies (Thabit et al., 2015; Kropff et al., 2015). Table 2 sum-
trol algorithm on a separate device, such as a smartphone, to ease de- marises the findings from key single- and dual-hormone randomised
velopment and regulatory issues (Fig. 5A). In September 2016, the US controlled free-living closed-loop studies. The longest randomised study
Food and Drug Regulatory Authority approved the first commercial by Thabit et al contrasted day-and-night hybrid closed-loop use with
closed-loop system (MiniMed® 670G pump, Medtronic, Northridge, CA) sensor-augmented pump therapy over three months and showed im-
for use in clinical practice, based on a non-randomised safety-focused provements in time in target glucose range, HbA1c and hypoglycaemia
pivotal clinical trial (FDA News Release, 2016). 670 G pump is a single- (Thabit et al., 2015).
hormone hybrid closed-loop system, with the control algorithm in- A meta-analysis including 585 adult and paediatric participants
tegrated in the pump (Fig. 5B). It is expected that 670G pump will be from 24 randomised controlled closed-loop studies showed that closed-
available in Europe in 2018/2019. Other groups and manufacturers are loop increased time spent in sensor glucose target range
also currently developing closed-loop systems for commercialisation. (3.9–10.0 mmol/l) by approximately 13 percentage points (p < .0001)
Bigfoot Smartloop™ system and the Omnipod Artificial Pancreas™ (Weisman et al., 2017). A comparison of time spent hypoglycaemic

Fig. 5. Panel A: FlorenceM hybrid closed-loop pro-

totype (University of Cambridge, Cambridge, UK)
using Android phone to host the control algorithm.
Panel B: MiniMed 670G hybrid closed-loop pump
with embedded control algorithm (Medtronic,
Northridge, CA, USA).

6
 L. Bally et al.

Table 2
Randomised controlled home studies of single-hormone and dual-hormone closed-loop systems.

Reference Study population N Intervention Comparator Study duration and setting Primary/co-primary outcome(s)

Single hormone closed-loop home studies

Hovorka et al., (2014) Children and 16 Single hormone, hybrid SAP Overnight x 3 weeks. Unsupervised % of time in sensor target range (3.9–8.0 mmol/l [70-144 mg/dl]): median 64%
adolescents closed-loop vs. 47%, CL vs. SAP (p < 0.0001)
Leelarathna et al., Adults 17 Single hormone, hybrid SAP Day-and-night x 7 days. Unsupervised % of time in target range (3.9–10 mmol/l [70–180 mg/dl]): median 75% vs.
(2014) closed-loop 62%, CL vs. SAP (p = 0.005).
Nimri et al., (2014) Adults and 19 Single hormone, hybrid SAP Overnight % of time below 3.9 mmol/l (70 mg/dl): median 3% vs. 5%, CL vs. SAP
adolescents closed-loop (p = 0.02).
Thabit et al., (2014) Adults 24 Single hormone, hybrid SAP Overnight x 4 weeks. Unsupervised % of time in target range (3.9–8.0 mmol/l [70–144 mg/dl]): mean 53% vs. 39%,
closed-loop CL vs. SAP (p = 0.0002).
(Nimri et al., (2014b) Adults and 15 Single hormone, hybrid SAP Overnight % of time below 3.9 mmol/l (70 mg/dl): median 4% vs. 49%, CL vs. SAP
adolescents closed-loop (p = 0.0034).
De Bock et al., (2015) Adults and 8 Single hormone, hybrid SAP Day-and-night x 5 days. Supervised % of time in sensor target range (4.0–9.9 mmol/l [72-178 mg/dl]): median
adolescents closed-loop 67.6% vs. 58.7%, CL vs. SAP + LGS (p = 0.233)
Thabit et al., (2015) Adolescents and 58 Single hormone, hybrid SAP Adults: Day-and-night x12weeks. % of time in sensor target range (adults 3.0–10 mmol/l [54-180 mg/dl]
adults closed-loop Adolescents: Overnight x 12 weeks. adolescents 3.9–8 mmol/l [70–144 mg/dl]): Adults –mean 67.7% vs. 56.8%, CL
Unsupervised vs. SAP (p < 0.001). Adolescents – mean 59.7% vs. 34.4% CL vs. SAP
(p = 0.004).
Kropff et al., (2015) Adults 32 Single hormone, hybrid SAP Dinnertime & overnight x 8 week. % of time in sensor target range (3.9–10 mmol/l [70–180 mg/dl]): mean 66.7%
closed-loop Unsupervised vs. 58.1% vs., CL vs. SAP (p < 0.001).
(Sharifi et al. (2016) Adolescents and 16 Single hormone, hybrid LGS Overnight x 4 night. Supervised % of time in sensor target range (4.0–8.0 mmol/l [72–144 mg/dl]): mean 59.4

7
adults closed-loop vs. 53.1%, CL vs. SAP + LGS (p = 0.14).
Tauschmann et al., Adolescents 12 Single hormone, hybrid SAP Day-and-night x 3 weeks. Unsupervised. % of time in target range (3.9–10 mmol/l [70–180 mg/dl]): mean 66.6% vs.
(2016) closed-loop 47.7%, CL vs. SAP (p < 0.001).
Bally et al., (2017) Adults 29 Single hormone, hybrid Usual pump therapy Day-and-night x 4 weeks. Unsupervised % of time in target range (3.9–10 mmol/l [70–180 mg/dl]): mean 76% vs. 66%.,
closed-loop CL vs. pump therapy(p < 0.001).
Forlenza et al., (2017) Adults 19 Single hormone, SAP Day-and-night x 2 weeks. Supervised % time with sensor glucose in target range (3.9–10 mmol/l [70-180 mg/dl]):
Hybrid closed-loop 71.6 vs. 65.2%; CL vs. SAP (p = 0.008)
Spaic et al., (2017) Adolescents and 30 Single-hormone PHHM PLGS Overnight x 42 nights % time with sensor glucose in target range (3.9–10 mmol/l [70-180mg/dl]): 78
adults vs. 71%, PHHM vs. PLGS nights (p < 0.001).

Dual-hormone closed-loop home studies

Haidar et al.,(2016b) Adolescents and 28 Dual-hormone, hybrid Single-hormone CL or SAP Overnight x 6 nights. Supervised % time with sensor glucose in target range (4.0–8.0 mmol/l [72-144 mg/
adults closed-loop dl]):median 81% vs. 76% vs. 47%, dual-CL vs. single-CL vs. SAP (p = 0.5 for dual-
CL vs. single CL, p < 0.001 for dual-CL vs. SAP, p < 0.001 for single-CL vs. SAP)
Blauw et al., (2016) Adults 16 Dual-hormone, fully Usual pump therapy 24h at CRF + day-and-night x 3 days at Median glucose: 7.3 mmol/l (132mg/dl) vs. 7.7 mmol/l (139 mg/dl), dual-CL vs.
closed-loop home. Supervised pump therapy (p = 0.123).
El-Khatib et al., Adults 43 Dual-hormone, hybrid Usual pump with blinded CGM Day-and-night x 11 days. Remotely Co-primary outcome (mean glucose and %time with sensor
(2017) closed-loop (own CGM use permitted if usual monitored glucose < 3.3 mmol/l (60 mg/dl)): mean glucose 7.8 mmol/l (141 mg/dl) vs.
care) 9.0 mmol/l(162mg/dl) (p < 0.0001), 0.6% vs. 1.9% (p < 0.0001), dual-CL vs.
pump therapy.
Haidar et al., (2017) Adults 23 Dual-hormone, hybrid Single-hormone CL SAP Day-and-night x 60h. Supervised %time with sensor glucose < 4.0 mmol/l (72 mg/dl): 3.6% vs. 3.9% vs. 7.9%,
closed-loop) dual-CL vs. single CL vs. SAP, p = 0.072 for dual-CL vs. single-CL, p = 0.002 for
dual-CL vs. SAP, p = 0.001 for single-CL vs. dual-CL) P < 0.017 considered
statistically significant

LGS, Low Glucose Suspend; PLGS, Predictive Low Glucose Suspend; PHHM, Predictive Hyperglycaemia and Hypoglycaemia Minimization; SAP, Sensor-Augmented Pump Therapy, CL, Closed-Loop.
International Journal of Pharmaceutics xxx (xxxx) xxx–xxx
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

(sensor glucose < 3.9 mmol/l) from 463 participants showed that need to be made aware of the need for associated activities, e.g. the
closed-loop reduced time spent hypoglycaemic by 2.5 percentage points need for sensor calibration, and bolusing for hybrid closed-loop sys-
or 35 min per day (p < .0001). tems, and troubleshooting resilience to resolve, for example, con-
The incremental benefits of a single-hormone closed-loop system nectivity problems.
were reported in a randomised crossover trial by Bally et al involving
type 1 diabetes adults with tight glycaemic control (HbA1c < 7.5% 5. Outlook and conclusion
[58 mmol/mol]). A hybrid closed-loop use in adults with baseline
HbA1c of 6.9% (52 mmol/mol) showed improvement in glucose control Closed-loop systems are unique through automated glucose-re-
whilst reducing the burden of hypoglycaemia (Bally et al., 2017). Im- sponsive insulin delivery at hyper- and hypoglycaemia. The FDA ap-
provements in time in target and lower mean glucose were also shown proval of the single-hormone hybrid closed-loop system has confirmed
using hybrid closed-loop in pregnant women with type 1 diabetes closed-loop systems as a realistic therapeutic modality for type 1 dia-
compared to sensor-augmented pump therapy (Stewart et al., 2016) betes.
without increasing the risk of hypoglycaemia. Strategies to improve closed-loop technology include the use faster-
The largest non-randomised closed-loop study to date by Bergenstal acting insulin formulations (Heise et al., 2017), local warming devices
et al involved 124 participants (30 adolescents aged 14–21 years and 94 such as the Insupad (El-Laboudi and Oliver, 2015), alternative insulin
adults aged 22–75 years) over 3 months with the primary aim being delivery routes (Dassau et al., 2017) and adjunctive incretin-based
safety rather than efficacy of closed-loop use (Bergenstal et al., 2016). therapies (Weinzimer et al., 2012; Sherr et al., 2016; Renukuntla et al.,
In the longest closed-loop study to date, also non-randomised, day-and- 2014). A further progress in glucose sensor technology may increase
night hybrid closed-loop was used in 14 adults over 6 months with sensor accuracy, prolong sensor lifetime and facilitate factory calibra-
median proportion of time in sensor glucose target range tion (Bolinder et al., 2016). The Eversense implantable sensor system
[3.9–10.0 mmol/l (70–180 mg/dl)] of 76% during daytime and 78% (Senseonics, Inc., MD, USA), which was recently approved in Europe,
during night-time. Closed-loop use declined over time from median allows long-term up to 6 months continuous use of sensor glucose with
values of 87% at the beginning to 70% at the end of the study consistent accuracy throughout its use (Kropff et al., 2017). The de-
(Kovatchev et al., 2017). velopment of stable glucagon formulations (Castle et al., 2016) and
The efficacy and safety of a single-hormone fully closed-loop system dual chamber pumps for insulin and glucagon co-delivery may provide
with intraperitoneal insulin delivery through the DiaPort system was incremental benefits to those with increased hypoglycaemia-burden
assessed in a non-randomised parallel design 24-h inpatient study by and risk. Minimising device burden through integration of closed-loop
Dassau et al involving 10 adults with type 1 diabetes (Dassau et al., components is likely to enhance user uptake. Cloud-based interface
2017). Compared to subcutaneous insulin delivery, intraperitoneal in- would enable users’ data to be uploaded and viewed by their healthcare
sulin delivery with a modified control algorithm led to higher percen- professionals, thereby facilitating virtual clinic consultations and re-
tage of time with plasma glucose in target range [4.4–7.7 mmol/l mote monitoring by parents and guardians. Personalisation of closed-
(79–139 mg/dl)] and lower mean plasma glucose (8.4 vs. 10.5 mmol/l loop systems can potentially be achieved by more adaptive algorithms
[151–189 mg/dl]) (p = 0.03 and p = 0.004). In spite of comparable (Doyle et al., 2014) and the use of input signals related to physical
time spent hypoglycaemic (2.5% vs. 4.1%, p = 0.42), those receiving activity and exercise such as heart rate, accelerometry (Jacobs et al.,
intraperitoneal insulin delivery required significantly higher amount of 2015).
rescue carbohydrates to prevent hypoglycaemia. Longer studies evaluating biomedical and psychosocial outcomes as
well as cost-effectiveness are needed to support reimbursement. The
4.5. Limitations immediate impact on existing structured education programmes is un-
known and adjustment may be needed. The rapid progress from the
Although a significant progress has been made, further enhance- bench to bedside has made closed-loop a realistic treatment option and
ments could improve closed-loop performance. The delayed absorption is expected to lead to better care and quality of life for people with type
of present rapid-acting insulin analogues remains a barrier to optimum 1 diabetes.
closed-loop performance (Home, 2012) and contributes to a less fa-
vourable performance during daytime compared to night-time as meal-
Author contributions
related excursions and physical activity induce rapid changes in insulin
requirements. As such, the delayed insulin action contributes to the
All authors designed the content. LB and HT wrote the manuscript.
residual risk of hypoglycaemia observed with single-hormone systems
All authors critically reviewed the report. No writing assistance was
in the late post-prandial period and during exercise. Ultrafast insulin
provided.
analogues with accelerated pharmacokinetics have been developed as a
solution to overcome these limitations (Heise et al., 2017). Another
approach to enhance the pharmacokinetic profile of subcutaneous in- Conflict of interest disclosures
sulin delivery comprises local warming devices which accelerates in-
sulin absorption following bolus delivery (El-Laboudi and Oliver, RH reports having received speaker honoraria from Eli Lilly and
2015). Novo Nordisk, serving on advisory panel for Eli Lilly and Novo Nordisk,
Although the addition of glucagon may confer additional benefits in receiving license fees from BBraun and Medtronic; and having served as
terms of hypoglycaemia reduction, it also adds to the complexity, cost, a consultant to BBraun. LB and HT declare no competing financial in-
and burden of the system to the user. Safety data on the long-term use terests exist.
of subcutaneous glucagon are lacking. A recent qualitative study as-
sessing the psychosocial impact of dual hormone closed-loop system in Funding/support
adults reported a reduction in diabetes-related distress, especially with
respect to hypoglycaemia. However, > 75% of users highlighted the LB received support from the Swiss National Science Foundation
burden of daily glucagon replacement and carrying multiple devices (P1BEP3_165297). Support for the Artificial Pancreas work by JDRF,
(Weissberg-Benchell et al., 2017). Device burden is recognised as an National Institute for Health Research Cambridge Biomedical Research
important factor potentially limiting the uptake and sustained use of Centre, Wellcome Trust Strategic Award (100574/Z/12/Z), EC Horizon
closed-loop systems, thus highlighting the importance of managing user 2020 (H2020-SC1-731560), NIDDK (DP3DK112176 and
expectations and needs (Iturralde et al., 2017). Closed-loop users will 1UC4DK108520-01).

8
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

References Metab. 101 (1), 214–223.

Haidar, A., Messier, V., Legault, L., Ladouceur, M., Rabasa-Lhoret, R., 2017. Outpatient
60-hour day-and-night glucose control with dual-hormone artificial pancreas, single-
Atkinson, M.A., Eisenbarth, G.S., Michels, A.W., 2014. Type 1 diabetes. Lancet 383 hormone artificial pancreas, or sensor-augmented pump therapy in adults with type 1
(9911), 69–82. diabetes: an open-label, randomised, crossover, controlled trial. Diabetes Obes.
Bally, L., Thabit, H., Hovorka, R., 2016. Finding the right route for insulin delivery – an Metab. 19 (5), 713–720.
overview of implantable pump therapy. Expert Opin. Drug. Deliv. 1–9. Heise, T., Stender-Petersen, K., Hövelmann, U., et al., 2017. Pharmacokinetic and phar-
Bally, L., Thabit, H., Kojzar, H., et al., 2017. Day-and-night glycaemic control with closed- macodynamic properties of faster-acting insulin aspart versus insulin aspart across a
loop insulin delivery versus conventional insulin pump therapy in free-living adults clinically relevant dose range in subjects with type 1 diabetes mellitus. Clin.
with well controlled type 1 diabetes: an open-label, randomised, crossover study. Pharmacokinet. 56 (6), 649–660.
Lancet Diabetes Endocrinol. 5 (4), 261–270. Hinshaw, L., Dalla Man, C., Nandy, D.K., et al., 2013. Diurnal pattern of insulin action in
Battelino, T., Nimri, R., Dovc, K., Phillip, M., Bratina, N., 2017. Prevention of hypogly- type 1 diabetes: implications for a closed-loop system. Diabetes 62 (7), 2223–2229.
cemia with predictive low glucose insulin suspension in children with type 1 diabetes: Home, P.D., 2012. The pharmacokinetics and pharmacodynamics of rapid-acting insulin
a randomized controlled trial. Diabetes Care 40 (6), 764–770. analogues and their clinical consequences. Diabetes Obes. Metab. 14 (9), 780–788.
Beato-Víbora, P., Yeoh, E., Rogers, H., Hopkins, D., Amiel, S.A., Choudhary, P., 2015. Hovorka, R., 2011. Closed-loop insulin delivery: from bench to clinical practice. Nat. Rev.
Sustained benefit of continuous subcutaneous insulin infusion on glycaemic control Endocrinol. 7 (7), 385–395.
and hypoglycaemia in adults with type 1 diabetes. Diabetes Med. 32 (11), Hovorka, R., Kumareswaran, K., Harris, J., et al., 2011. Overnight closed loop insulin
1453–1459. delivery (artificial pancreas) in adults with type 1 diabetes: crossover randomised
Beck, R.W., Riddlesworth, T.D., Ruedy, K.J., et al., 2017. Effect of initiating use of an controlled studies. BMJ 342, d1855.
insulin pump in adults with type 1 diabetes using multiple daily insulin injections and Hovorka, R., Elleri, D., Thabit, H., et al., 2014. Overnight closed-loop insulin delivery in
continuous glucose monitoring (DIAMOND): a multicentre, randomised controlled young people with type 1 diabetes: a free-living, randomized clinical trial. Diabetes
trial. Lancet Diabetes Endocrinol. 5 (9), 700–708. Care 37 (5), 1204–1211.
Bergenstal, R.M., Klonoff, D.C., Garg, S.K., et al., 2013. Threshold-based insulin-pump Iturralde, E., Tanenbaum, M.L., Hanes, S.J., et al., 2017. Expectations and attitudes of
interruption for reduction of hypoglycemia. N. Engl. J. Med. 369 (3), 224–232. individuals with type 1 diabetes after using a hybrid closed loop system. Diabetes
Bergenstal, R.M., Garg, S., Weinzimer, S.A., et al., 2016. Safety of a hybrid closed-loop Educ. 43 (2), 223–232.
insulin delivery system in patients with type 1 diabetes. JAMA 316 (13), 1407–1408. Jacobs, P.G., Resalat, N., El Youssef, J., et al., 2015. Incorporating an exercise detection,
Blauw, H., Keith-Hynes, P., Koops, R., DeVries, J.H., 2016. A review of safety and design grading, and hormone dosing algorithm into the artificial pancreas using accel-
requirements of the artificial pancreas. Ann. Biomed. Eng. 44 (11), 3158–3172. erometry and heart rate. J. Diabetes Sci. Technol. 9 (6), 1175–1184.
Bolinder, J., Antuna, R., Geelhoed-Duijvestijn, P., Kröger, J., Weitgasser, R., 2016. Novel Karges, B., Schwandt, A., Heidtmann, B., et al., 2017. Association of insulin pump therapy
glucose-sensing technology and hypoglycaemia in type 1 diabetes: a multicentre, vs insulin injection therapy with severe hypoglycemia, ketoacidosis, and glycemic
non-masked, randomised controlled trial. Lancet. http://dx.doi.org/10.1016/S0140- control among children, adolescents, and young adults with type 1 diabetes. JAMA
6736(16)31535-5. 318 (14), 1358–1366.
Buckingham, B.A., Cameron, F., Calhoun, P., et al., 2013. Outpatient safety assessment of Kilpatrick, E.S., Rigby, A.S., Goode, K., Atkin, S.L., 2007. Relating mean blood glucose
an in-home predictive low-glucose suspend system with type 1 diabetes subjects at and glucose variability to the risk of multiple episodes of hypoglycaemia in type 1
elevated risk of nocturnal hypoglycemia. Diabetes Technol. Ther. 15 (8), 622–627. diabetes. Diabetologia 50 (12), 2553–2561.
Buckingham, B.A., Raghinaru, D., Cameron, F., et al., 2015. Predictive low-glucose insulin Kovatchev, B., Cheng, P., Anderson, S.M., et al., 2017. Feasibility of long-term closed-loop
suspension reduces duration of nocturnal hypoglycemia in children without in- control: A multicenter 6-month trial of 24/7 automated insulin delivery. Diabetes
creasing ketosis. Diabetes Care 38 (7), 1197–1204. Technol. Ther. 19 (1), 18–24.
Caputo, N., Castle, J.R., Bergstrom, C.P., et al., 2013. Mechanisms of glucagon de- Kropff, J., Del Favero, S., Place, J., et al., 2015. 2 month evening and night closed-loop
gradation at alkaline pH. Peptides 45, 40–47. glucose control in patients with type 1 diabetes under free-living conditions: a ran-
Castle, J.R., Youssef, J.E., Branigan, D., et al., 2016. Comparative pharmacokinetic/ domised crossover trial. Lancet Diabetes Endocrinol. 3 (12), 939–947.
pharmacodynamic study of liquid stable glucagon versus lyophilized glucagon in type Kropff, J., Choudhary, P., Neupane, S., et al., 2017. Accuracy and longevity of an im-
1 diabetes subjects. J. Diabetes Sci. Technol. 10 (5), 1101–1107. http://dx.doi.org/ plantable continuous glucose sensor in the precise study: a 180-day, prospective,
10.1177/1932296816653141. multicenter, pivotal trial. Diabetes Care 40 (1), 63–68.
Cummins, E., Royle, P., Snaith, A., et al., 2010. Clinical effectiveness and cost-effective- Leelarathna, L., Dellweg, S., Mader, J.K., et al., 2014. Day and night home closed-loop
ness of continuous subcutaneous insulin infusion for diabetes: systematic review and insulin delivery in adults with type 1 diabetes: three-center randomized crossover
economic evaluation. Health Technol. Assess. 14 (11), 1–181 iii-iv, xi-xvi,. study. Diabetes Care 37 (7), 1931–1937.
Dassau, E., Renard, E., Place, J., et al., 2017. Intraperitoneal insulin delivery provides Liebl, A., Frei, T., 2003. Safety in continuous intraperitoneal insulin infusion CIPII via
superior glycemic regulation to subcutaneous insulin delivery in model predictive DiaPort (R): results from the DiaPort-001-study. Diabetologia 46, A46–A47.
control-based fully-automated artificial pancreas in patients with type 1 diabetes: a Liebl, A., Hoogma, R., Renard, E., et al., 2009. A reduction in severe hypoglycaemia in
pilot study. Diabetes Obes. Metab. 19 (12), 1698–1705. type 1 diabetes in a randomized crossover study of continuous intraperitoneal com-
de Bock, M.I., Roy, A., Cooper, M.N., et al., 2015. Feasibility of outpatient 24-hour closed- pared with subcutaneous insulin infusion. Diabetes Obes. Metab. 11 (11),
loop insulin delivery. Diabetes Care 38 (11), e186–e187. 1001–1008.
Del Favero, S., Place, J., Kropff, J., et al., 2015. Multicenter outpatient dinner/overnight Logtenberg, S.J., Kleefstra, N., Houweling, S.T., et al., 2009. Improved glycemic control
reduction of hypoglycemia and increased time of glucose in target with a wearable with intraperitoneal versus subcutaneous insulin in type 1 diabetes: a randomized
artificial pancreas using modular model predictive control in adults with type 1 controlled trial. Diabetes Care 32 (8), 1372–1377.
diabetes. Diabetes Obes. Metab. 17 (5), 468–476. Ly, T.T., Nicholas, J.A., Retterath, A., Lim, E.M., Davis, E.A., Jones, T.W., 2013. Effect of
DeVries, J.H., Snoek, F.J., Kostense, P.J., Masurel, N., Heine, R.J., Group, D.I.P.S., 2002. A sensor-augmented insulin pump therapy and automated insulin suspension vs stan-
randomized trial of continuous subcutaneous insulin infusion and intensive injection dard insulin pump therapy on hypoglycemia in patients with type 1 diabetes: a
therapy in type 1 diabetes for patients with long-standing poor glycemic control. randomized clinical trial. JAMA 310 (12), 1240–1247.
Diabetes Care 25 (11), 2074–2080. Ly, T.T., Brnabic, A.J., Eggleston, A., et al., 2014. A cost-effectiveness analysis of sensor-
Doyle, F.J., Huyett, L.M., Lee, J.B., Zisser, H.C., Dassau, E., 2014. Closed-loop artificial augmented insulin pump therapy and automated insulin suspension versus standard
pancreas systems: engineering the algorithms. Diabetes Care 37 (5), 1191–1197. pump therapy for hypoglycemic unaware patients with type 1 diabetes. Value Health
El-Khatib, F.H., Balliro, C., Hillard, M.A., et al., 2017. Home use of a bihormonal bionic 17 (5), 561–569.
pancreas versus insulin pump therapy in adults with type 1 diabetes: a multicentre Maahs, D.M., Calhoun, P., Buckingham, B.A., et al., 2014. A randomized trial of a home
randomised crossover trial. Lancet 389 (10067), 369–380. system to reduce nocturnal hypoglycemia in type 1 diabetes. Diabetes Care 37 (7),
El-Laboudi, A.H., Oliver, N., 2015. Towards a physiological prandial insulin profile: en- 1885–1891.
hancement of subcutaneously injected prandial insulin using local warming devices. Misso, M.L., Egberts, K.J., Page, M., O'Connor, D., Shaw, J., 2010. Continuous sub-
Diabetes Ther. 6 (3), 257–272. cutaneous insulin infusion (CSII) versus multiple insulin injections for type 1 diabetes
Fazeli Farsani, S., Brodovicz, K., Soleymanlou, N., Marquard, J., Wissinger, E., Maiese, mellitus. Cochrane Database Syst. Rev. (1), CD005103.
B.A., 2017. Incidence and prevalence of diabetic ketoacidosis (DKA) among adults Nathan, D.M., Genuth, S., Lachin, J., Cleary, P., Crofford, O., Davis, M., Rand, L., Siebert,
with type 1 diabetes mellitus (T1D): a systematic literature review. BMJ Open 7 (7), C., 1993. The effect of intensive treatment of diabetes on the development and
e016587. progression of long-term complications in insulin-dependent diabetes mellitus. The
FDA News Release, 2016. FDA Approves First Automated Insulin Delivery Device Fortype diabetes control and complications trial research group. N. Engl. J. Med. 329 (14),
1 Diabetes. http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ 977–986.
ucm522974.htm, October. Nimri, R., Muller, I., Atlas, E., et al., 2014a. MD-logic overnight control for 6 weeks of
Forlenza, G.P., Deshpande, S., Ly, T.T., et al., 2017. Application of zone model predictive home use in patients with type 1 diabetes: randomized crossover trial. Diabetes Care
control artificial pancreas during extended use of infusion set and sensor: a rando- 37 (11), 3025–3032.
mized crossover-controlled home-use trial. Diabetes Care 40 (8), 1096–1102. Nimri, R., Muller, I., Atlas, E., et al., 2014b. Night glucose control with MD-logic artificial
Frier, B.M., 2014. Hypoglycaemia in diabetes mellitus: epidemiology and clinical im- pancreas in home setting: a single blind, randomized crossover trial-interim analysis.
plications. Nat. Rev. Endocrinol. 10 (12), 711–722. Pediatr. Diabetes 15 (2), 91–99.
Haidar, A., Smaoui, M.R., Legault, L., Rabasa-Lhoret, R., 2016a. The role of glucagon in Patel, P.J., Benasi, K., Ferrari, G., et al., 2014. Randomized trial of infusion set function:
the artificial pancreas. Lancet Diabetes Endocrinol. 4 (6), 476–479. http://dx.doi. steel versus teflon. Diabetes Technol. Ther. 16 (1), 15–19.
org/10.1016/S2213-8587(16)30006-7. Peters, A.L., Ahmann, A.J., Battelino, T., et al., 2016. Diabetes technology-continuous
Haidar, A., Rabasa-Lhoret, R., Legault, L., et al., 2016b. Single- and dual-hormone arti- subcutaneous insulin infusion therapy and continuous glucose monitoring in adults:
ficial pancreas for overnight glucose control in type 1 diabetes. J. Clin. Endocrinol. an endocrine society clinical practice guideline. J. Clin. Endocrinol. Metab. 101 (11),

9
L. Bally et al. International Journal of Pharmaceutics xxx (xxxx) xxx–xxx

3922–3937. Spaic, T., Driscoll, M., Raghinaru, D., et al., 2017. Predictive hyperglycemia and hy-
Phillip, M., Battelino, T., Atlas, E., et al., 2013. Nocturnal glucose control with an artificial poglycemia minimization: in-home evaluation of safety, feasibility, and efficacy in
pancreas at a diabetes camp. N. Engl. J. Med. 368 (9), 824–833. overnight glucose control in type 1 diabetes. Diabetes Care 40 (3), 359–366.
Pickup, J.C., 2012. Management of diabetes mellitus: is the pump mightier than the pen? Stewart, Z.A., Wilinska, M.E., Hartnell, S., et al., 2016. Closed-loop insulin delivery
Nat. Rev. Endocrinol. 8 (7), 425–433. during pregnancy in women with type 1 diabetes. N. Engl. J. Med. 375 (7), 644–654.
Pickup, J.C., Sutton, A.J., 2008. Severe hypoglycaemia and glycaemic control in type 1 Taleb, N., Haidar, A., Messier, V., Gingras, V., Legault, L., Rabasa-Lhoret, R., 2017.
diabetes: meta-analysis of multiple daily insulin injections compared with continuous Glucagon in artificial pancreas systems: potential benefits and safety profile of future
subcutaneous insulin infusion. Diabetes Med. 25 (7), 765–774. chronic use. Diabetes Obes. Metab. 19 (1), 13–23.
Pickup, J.C., Keen, H., Parsons, J.A., Alberti, K.G., 1978. Continuous subcutaneous insulin Tauschmann, M., Allen, J.M., Wilinska, M.E., et al., 2016. Home use of day-and-night
infusion: an approach to achieving normoglycaemia. Br. Med. J. 1 (6107), 204–207. hybrid closed-loop insulin delivery in suboptimally controlled adolescents with type
Pickup, J.C., Freeman, S.C., Sutton, A.J., 2011. Glycaemic control in type 1 diabetes 1 diabetes: a 3-week, free-living, randomized crossover trial. Diabetes Care 39 (11),
during real time continuous glucose monitoring compared with self monitoring of 2019–2025.
blood glucose: meta-analysis of randomised controlled trials using individual patient Thabit, H., Hovorka, R., 2016. Coming of age: the artificial pancreas for type 1 diabetes.
data. BMJ 343, d3805. Diabetologia 59 (9), 1795–1805.
Pickup, J.C., Yemane, N., Brackenridge, A., Pender, S., 2014. Nonmetabolic complications Thabit, H., Lubina-Solomon, A., Stadler, M., et al., 2014. Home use of closed-loop insulin
of continuous subcutaneous insulin infusion: a patient survey. Diabetes Technol Ther. delivery for overnight glucose control in adults with type 1 diabetes: a 4-week,
16 (3), 145–149. multicentre, randomised crossover study. Lancet Diabetes Endocrinol. 2 (9),
Pozzilli, P., Battelino, T., Danne, T., Hovorka, R., Jarosz-Chobot, P., Renard, E., 2015. 701–709.
Continuous subcutaneous insulin infusion in diabetes: patient populations, safety, Thabit, H., Tauschmann, M., Allen, J.M., et al., 2015. Home use of an artificial beta cell in
efficacy, and pharmacoeconomics. Diabetes Metab. Res. Rev.(April). http://dx.doi. type 1 diabetes. N. Engl. J. Med. 373 (22), 2129–2140.
org/10.1002/dmrr.2653. Trevitt, S., Simpson, S., Wood, A., 2016. Artificial pancreas device systems for the closed-
Renukuntla, V.S., Ramchandani, N., Trast, J., Cantwell, M., Heptulla, R.A., 2014. Role of loop control of type 1 diabetes: what systems are in development? J. Diabetes Sci.
glucagon-like peptide-1 analogue versus amylin as an adjuvant therapy in type 1 Technol. 10 (3), 714–723. http://dx.doi.org/10.1177/1932296815617968.
diabetes in a closed loop setting with ePID algorithm. J. Diabetes Sci. Technol 8 (5), Underland, L.J., Ilkowitz, J.T., Katikaneni, R., Dowd, A., Heptulla, R.A., 2017. Use of
1011–1017. sitagliptin with closed-loop technology to decrease postprandial blood glucose in type
Rodbard, D., 2016. Continuous glucose monitoring: a review of successes, challenges, and 1 diabetes. J. Diabetes Sci. Technol. 11 (3), 602–610. http://dx.doi.org/10.1177/
opportunities. Diabetes Technol. Ther. 18 (Suppl. 2), S23–S213. 1932296817699847.
Roze, S., Smith-Palmer, J., Valentine, W., de Portu, S., Nørgaard, K., Pickup, J.C., 2015. van Dijk, P.R., Logtenberg, S.J., Hendriks, S.H., et al., 2015. Intraperitoneal versus sub-
Cost-effectiveness of continuous subcutaneous insulin infusion versus multiple daily cutaneous insulin therapy in the treatment of type 1 diabetes mellitus. Neth. J. Med.
injections of insulin in type 1 diabetes: a systematic review. Diabetes Med. 32 (11), 73 (9), 399–409.
1415–1424. Weinzimer, S.A., Steil, G.M., Swan, K.L., Dziura, J., Kurtz, N., Tamborlane, W.V., 2008.
Roze, S., Smith-Palmer, J., Valentine, W., et al., 2016. Cost-effectiveness of sensor-aug- Fully automated closed-loop insulin delivery versus semiautomated hybrid control in
mented pump therapy with low glucose suspend versus standard insulin pump pediatric patients with type 1 diabetes using an artificial pancreas. Diabetes Care 31
therapy in two different patient populations with type 1 diabetes in France. Diabetes (5), 934–939.
Technol. Ther. 18 (2), 75–84. Weinzimer, S.A., Sherr, J.L., Cengiz, E., et al., 2012. Effect of pramlintide on prandial
Ruan, Y., Thabit, H., Leelarathna, L., et al., 2016. Variability of insulin requirements over glycemic excursions during closed-loop control in adolescents and young adults with
12 weeks of closed-loop insulin delivery in adults with type 1 diabetes. Diabetes Care type 1 diabetes. Diabetes Care 35 (10), 1994–1999.
39 (5), 830–832. Weisman, A., Bai, J.W., Cardinez, M., Kramer, C.K., Perkins, B.A., 2017. Effect of artificial
Schade, D.S., Eaton, R.P., Davis, T., et al., 1981. The kinetics of peritoneal insulin ab- pancreas systems on glycaemic control in patients with type 1 diabetes: a systematic
sorption. Metabolism 30 (2), 149–155. review and meta-analysis of outpatient randomised controlled trials. Lancet Diabetes
Schaepelynck, P., Renard, E., Jeandidier, N., et al., 2011. A recent survey confirms the Endocrinol.
efficacy and the safety of implanted insulin pumps during long-term use in poorly Weissberg-Benchell, J., Antisdel-Lomaglio, J., Seshadri, R., 2003. Insulin pump therapy: a
controlled type 1 diabetes patients. Diabetes Technol. Ther. 13 (6), 657–660. meta-analysis. Diabetes Care 26 (4), 1079–1087.
Sharifi, A., De Bock, M.I., Jayawardene, D., et al., 2016. Glycemia, treatment satisfaction, Weissberg-Benchell, J., Hessler, D., Fisher, L., Russell, S.J., Polonsky, W.H., 2017. Impact
cognition, and sleep quality in adults and adolescents with type 1 diabetes when of an automated bihormonal delivery system on psychosocial outcomes in adults with
using a closed-loop system overnight versus sensor-augmented pump with low-glu- type 1 diabetes. Diabetes Technol. Ther.
cose suspend function: a randomized crossover study. Diabetes Technol. Ther. 18 Zhong, A., Choudhary, P., McMahon, C., et al., 2016. Effectiveness of automated insulin
(12), 772–783. management features of the MiniMed(®) 640G sensor-augmented insulin pump.
Sherr, J.L., Patel, N.S., Michaud, C.I., et al., 2016. Mitigating meal-related glycemic ex- Diabetes Technol. Ther. 18 (10), 657–663.
cursions in an insulin-sparing manner during closed-loop insulin delivery: the ben- Zisser, H., Robinson, L., Bevier, W., et al., 2008. Bolus calculator: a review of four "smart"
eficial effects of adjunctive pramlintide and liraglutide. Diabetes Care 39 (7), insulin pumps. Diabetes Technol. Ther. 10 (6), 441–444.
1127–1134.

10