==> picture [215 x 22] intentionally omitted <==

Multiple Technology Appraisal

Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes

Committee papers

==> picture [215 x 22] intentionally omitted <==

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

MULTIPLE TECHNOLOGY APPRAISAL

Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes

Background:

Following the scoping workshop on 23 February 2021, stakeholders were informed that NICE was considering transferring the assessment from diagnostic guidance (DG) process, to a multiple technology assessment (MTA) process. This was then confirmed, therefore the assessment was transferred on to the MTA pathway and will publish as technology appraisal guidance rather than diagnostics guidance.

The assessment was paused in July 2021 to allow real world data to be collected by NHS England and NHS Improvement on the use of hybrid closed loop systems for people with type 1 diabetes in the NHS to be included in the assessment. During this time the recommendations on glucose monitoring in the NICE guideline on Type 1 diabetes in adults: diagnosis and management (NG17) were revised, the final scope and protocol were updated to reflect this.

Contents:

• 1 The final scope and list of stakeholders are available via the hyperlinks

• 2 Full preceding guidance: Integrated sensor-augmented pump therapy systems for managing blood glucose levels in type 1 diabetes (the MiniMed Paradigm Veo system and the Vibe and G4 PLATINUM CGM system) [DG21]: https://www.nice.org.uk/guidance/dg21

3 Overview

• 4 Assessment Report and appendices (September 2022) prepared by Warwick Evidence, Warwick Medical School, University of Warwick (Note: this report has been superseded by paper 10 - Updated Assessment Report (15 November 2022))

• 5 Consultee, commentator and expert comments on the Assessment Report (see list below) and responses to the comments prepared by Warwick Evidence :

  • Air Liquide Healthcare

  • Dexcom

  • Medtronic

  • Tandem Diabetes Care Inc

  • Insulet

  • Diabetes UK

  • JDFR

  • ABC Diabetes Technology Network UK

  • NHS England

  • Sufyan Hussain (expert)

  • Fiona Regan (expert)

==> picture [215 x 22] intentionally omitted <==

6 Company submission summaries from:

• a. Camdiab Ltd

• b. Dexcom International Ltd

• c. Medtronic Ltd

• d. Tandem Diabetes Care Inc

(Air Liquide Healthcare – not providing a submission)

7 NHS England - Hybrid Closed Loop Systems pilot data:

• Adult centres

• Paediatric centres

• Enablers and barriers affecting adoption of diabetic technologies

8 Patient and professional groups submissions from:

• Diabetes UK

• JDRF

9 Expert personal perspectives from:

Experts appointed via the MTA expert nomination process:

• Julie Brake – clinical expert, nominated by Insulet International Ltd

• Joanne Richardson - patient expert, nominated by National Children and Young People’s Diabetes Network

Experts appointed by the Diagnostics Assessment Programme

(Originally appointed as specialist committee members, before the topic was re-routed as a Multiple Technology Appraisal):

• Nicola Birchmore - clinical expert

• Alison Finney - patient expert

• Jeff Foot - patient expert

• Peter Hindmarsh - clinical expert

• Sufyan Hussain - clinical expert

• Fiona Regan - clinical expert

• Philip Weston - clinical expert

10 Updated Assessment Report (15 November 2022) prepared by Warwick Evidence, Warwick Medical School, University of Warwick

Any information supplied to NICE which has been marked as confidential has been redacted. All personal information has also been redacted.

Evidence overview: Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes

This overview summarises the main issues the diagnostics advisory committee needs to consider. It should be read together with the final scope and the updated external assessment report (15 November 2022).

1 Aims and scope

The purpose of this assessment is to evaluate the clinical and cost effectiveness of using hybrid closed loop systems for managing glucose levels in type 1 diabetes.

In type 1 diabetes, a person’s blood glucose level becomes too high (hyperglycaemia) because there is no, or very little, production of insulin by the pancreas. The goal of treatment in type 1 diabetes is to keep blood glucose within a healthy range by providing the body with supplemental insulin. If the level of circulating insulin becomes too high, blood glucose levels can become too low leading to hypoglycaemia (also known as a hypo).

The management of type 1 diabetes has several components and typically involves lifestyle adjustments, regular measuring of blood glucose levels, use of multiple daily insulin injections or continuous subcutaneous insulin infusion (CSII) and periodic assessment of blood glucose control. Long ‑ term monitoring of blood glucose control can be done by measuring glycated haemoglobin (HbA1c levels), which is the average plasma glucose over the preceding 3 months. NICE guidelines on diabetes (type 1 and type 2) in children and young people, type 1 diabetes in adults and diabetes in pregnancy recommend that people with type 1 diabetes should aim for a target HbA1c level of 48 millimoles per mole (6.5%) or lower to minimise the risk of long term complications from diabetes.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 1 of 54

Time in range is a measure of glycaemic control which shows the percentage of time a person spends within a target glucose range. It is obtained from continuous glucose monitor data and gives an idea of changes in glucose patterns within a day and between days. The international consensus on time in range recommends a time in range of at least 70% in a glucose range of 3.9 to 10 millimoles per litre for people with type 1 diabetes. Time below range (percentage of time between 3.0 to 3.9 mmol/litre) is associated with increased risk of severe hypoglycaemia, while time above range (percentage of time between 10 to 13.9 mmol/litre) may indicate a risk of ketoacidosis. Blood glucose monitoring can be done by self monitoring (capillary blood testing), or by real time continuous (rtCGM) or intermittently scanned continuous glucose monitors (isCGM). The Diabetes UK position statement on the appropriate use of technology in type 1 diabetes is shown in figure 1.

Figure 1 Technology care pathway for type 1 diabetes

==> picture [417 x 267] intentionally omitted <==

Adapted from Type 1 diabetes technology pathway: consensus statement for the use of technology in Type 1 diabetes Choudhary et al. 2019.

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 2 of 54

Integrated sensor augmented pump (SAP) systems combine rtCGM with continuous subcutaneous insulin infusion (CSII). This assessment is an update of NICE diagnostics guidance 21 (DG21). This guidance assessed 2 integrated SAP systems and recommended the MiniMed Paradigm Veo system as an option for managing blood glucose levels in people with type 1 diabetes if they have episodes of disabling hypoglycaemia despite optimal management with continuous subcutaneous insulin infusion. The Vibe and G4 PLATINUM CGM system was not recommended for routine adoption by the NHS as further evidence was needed to show the clinical effectiveness. The SAP systems assessed in DG21 are no longer available to the NHS and they have been replaced by successor systems with enhanced features.

Hybrid closed loop (HCL) systems use a mathematical algorithm to automatically drive insulin delivery in response to continuously monitored interstitial fluid glucose levels. They use a combination of real-time glucose monitoring from a CGM device and a control algorithm to direct insulin delivery through a CSII pump. Basal insulin is delivered automatically whereas bolus doses at mealtimes are manually delivered by the user. Some of these systems are built by combining interoperable devices from different manufacturers.

Decision question

Does the use of hybrid closed loop systems for managing glucose levels in type 1 diabetes represent a clinically and cost-effective use of NHS resources?

Populations

People with type 1 diabetes who are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous subcutaneous insulin infusion or real time continuous glucose monitoring or intermittently scanned glucose monitoring. These difficulties may include:

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 3 of 54

• not maintaining HbA1c levels of 6.5% or below or

• not maintaining at least 70% time in range of 3.9 -10 mmol/litre or

• ongoing disabling hypoglycaemia

If evidence permits the following subpopulations should be included:

• Women with type 1 diabetes who are pregnant and those planning pregnancy (not including gestational diabetes). Please note that in this assessment this subpopulation is not required to fulfil the criteria of prior use of at least 1 technology.

• Children with type 1 diabetes. If possible, evidence should be analysed based on the following age groups:

  • 5 years and under

  • 6 to 11 years

  • 12 to 19 years

• People with extreme fear of hypoglycaemia

• People with diabetes related complications that are at risk of deterioration

Interventions

Hybrid closed loop systems

Comparators

For the economic modelling the comparators will be:

• Real time continuous glucose monitoring with continuous subcutaneous insulin infusion (non-integrated)

• Intermittently scanned glucose monitoring with continuous subcutaneous insulin infusion

Where evidence permits scenarios assessing the following comparators should be presented for women with type 1 diabetes who are pregnant or those planning pregnancy:

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 4 of 54

• Real time continuous glucose monitoring with multiple daily insulin injections

• Intermittently scanned glucose monitoring with multiple daily insulin injections

• Self blood glucose monitoring with continuous subcutaneous insulin infusion

Healthcare setting

The healthcare setting for the interventions is self-use supervised by primary or secondary care. Figure 2 shows how the population, intervention and comparator fit into the care pathway. Further details, including descriptions of the interventions, comparator, care pathway and outcomes, are in the final scope for Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes.

Figure 2 Overview of population and technologies in the current care pathway

==> picture [447 x 249] intentionally omitted <==

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 5 of 54

2 Clinical effectiveness evidence

The EAG did a systematic review to identify evidence on the clinical effectiveness and diagnostic accuracy of hybrid closed loop systems for managing blood glucose levels in type 1 diabetes. Find the full systematic review results from page 66 of the external assessment report.

Overview of included studies

Randomised controlled trials

There were 12 randomised controlled trials (RCTs) (11 publications). Most were multinational trials with participants recruited from centres in various countries including Australia, Austria, France, Germany, Israel, Luxembourg, New Zealand, Slovenia, UK and USA. The interventions varied across the different RCTs and some used systems consisting of interoperable devices from different manufacturers. Thabit et al. 2015 reported the results of 2 RCTs (1 in adults and 1 in children and adolescents). The study by Collyns et al. 2021 reported 3 separate sets of results from 1 RCT (children, adolescents and adults). Table 1 shows the population characteristics of the RCTs. Find more details of the included RCTs, including details of HCL systems used and comparators, in appendix 2 (page 287) of the external assessment report.

The EAG said that the inclusion criteria used in the RCTs were relatively narrow and most participants had reasonably good glycaemic control at entry. The EAG said that overall, studies were heterogeneous in terms of RCT design (parallel groups or cross over design with wash-out phase between different treatments), population, participants age, gender, numbers of participants and other demographics including run-in times, duration of observation periods, and number and types of previous treatments. The EAG also said that the studies did not consistently describe comparators.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 6 of 54

Table 1 Population characteristics of the included RCTs

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 7 of 54

LGS/PLGS = low glucose suspend/predictive low glucose suspend

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 8 of 54

RCTs in which HCL treatment was received for 4 or more weeks (range of 4 to 26 weeks) were included if the comparator was relevant to the decision problem. Comparators were SAP systems classified as continuous subcutaneous insulin infusion (CSII) plus continuous glucose monitoring (CGM) and low glucose suspend or predictive low glucose suspend (LGS/PLGS) systems. Where reported, CGMs used in the studies were rtCGMs. The LGS/PLGS systems are earlier versions of automated insulin delivery systems that use continuous glucose sensor data to allow immediate real ‑ time manual adjustment of insulin therapy. The systems produce alerts if the glucose levels become too high or too low, if levels are rapidly changing, or if the system predicts that levels will be too high or too low in the near future. LGS systems can automatically suspend insulin delivery if there is no ‑ response to a low glucose warning. PLGS systems automatically suspend insulin if the system predicts that the person is heading towards hypoglycaemia. Other insulin adjustments are made manually by the user. LGS and PLGS systems are no longer available in the NHS.

Table 3 in the external assessment report summarises the baseline characteristics and the main outcome measures reported in the RCTs (pages 73 to 79).

Quality assessment of RCTs

The EAG used the revised Cochrane risk of bias tool for randomised trials to critically appraise the 12 RCTs. In this assessment the EAG treated the 2 RCTs included in Thabit et al. 2015 as 1 study. The EAG said that 5 of the RCTs had some concerns about their risk of bias and 3 had a high risk of bias (Benhamou et al. 2021, von dem Berge et al. 2022 and Collyns et al. 2021). High risk of bias was most common in relation to the randomisation process and deviations from intended interventions. In terms of randomisation 1 RCT (Collyns et al. 2021) had a high risk of bias and 4 had some concerns. In terms of deviations from intended interventions 1 RCT had a high risk of bias (Benhamou et al. 2022) and 6 had some concerns. Three RCTs also had

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 9 of 54

some concerns of the risk of bias relating to selection of the reported results. All 12 RCTs had a low risk of bias in relation to both missing outcome data and outcomes measurement. Full details of the quality assessment of the RCTs are on pages 109 to 110 of the external assessment report.

Observational studies

Nine observational studies were identified that provided outcomes indicating glycaemic performance in people with type 1 diabetes mellitus (T1DM), using HCL or AHCL (advanced HCL) systems. Two of the observational studies were NHSE pilot studies, 1 in adults and 1 in children and young people (CYP). The adult study included 570 adults with T1DM (with complete followup data) from 31 diabetes centres across England that started HCL therapy. The CYP study included 251 children and young people (under 19 years), with T1DM for at least a year and had 2 HbA1c measures prior to the start of HCL. Most observational studies used similar inclusion criteria to those used in the RCTs. The EAG said that the NHSE pilot studies were broader in recruitment and included adult participants that had poorer glycaemic control in terms of HbA1c and hyperglycaemia at baseline than the other observational studies.

The observational studies included more participants than the RCTs. For the NHSE pilot data, the adult study accumulated over 200 person years of HCL observations and the CYP study around 100 person years.

Details of the population characteristics of the 9 observational studies are in table 4 on page 90 in the external assessment report. Outcome results reported in the observational studies are shown in table 5, pages 92 to 97.

No quality assessment was done on the observational studies.

Quality assessment of NHS England evidence

The EAG said that the NHSE pilot studies were non-randomised studies with no control group and with a before-after study design. It said that the beforeand-after study design limited the scientific value of the evidence because

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 10 of 54

there was a greater risk of bias due to lack of randomisation, lack of a true control, and selection bias. In addition, the findings of the two pilots are interim results and therefore may not give the full results.

Intermediate outcomes: RCTs

All RCT studies reported results for change in percentage HbA1c, change in percentage time in range indicating satisfactory glycaemic control (3.9 to 10 or 3.5 to 7.8 mmol/litre, percentage time above range (more than 10 mmol/litre), and percentage time below range (less than 3.9, 3.5, 3.3, 3.0 or 2.8 mmol/litre depending on the study). The following outcome sections for the RCTs include descriptive comparisons with the NHSE pilot study data where relevant. Because LGS/PLGS systems are no longer available to purchase in the UK, results comparing the clinical effectiveness of CSII plus CGM with LGS/PLGS are not discussed in the following sections.

Change in HbA1c percentage

A reduction of HbA1c over time indicates improved glycaemic control. A negative mean difference or net effect size estimate (ES) comparing HCL with the comparator indicates superior glycaemic control with HCL. The study by Kariyawasam et al. was not included because it only reported baseline data so change in HbA1c could not be estimated and the net effect was not reported. Stewart et al. was not included because it only reported end of study medians (no baseline) so change could not be estimated. Figure 3 shows the change from baseline in percentage HbA1c for each arm over the treatment period for the different RCTs.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 11 of 54

Figure 3 Change (mean ± sd or median) in percentage HbA1c over treatment period in RCTs

==> picture [417 x 302] intentionally omitted <==

----- Start of picture text -----

----- End of picture text -----

Weeks = treatment period; BL = baseline value; comp = comparator; HCL = hybrid closed loop; N = number of participants; yr = years; ES = net effect size mean difference 95% CI [HCL vs. comparator]; medians have no error bars. For Collyns et al. ES was only reported for all 3 age groups combined.

The EAG said that the range of mean baseline percentage HbA1c in the RCTs was narrow (7.4 to 8.3). In all studies, the reduction in percentage HbA1c was greater for HCL than the comparator. Change in percentage HbA1c over the treatment period in HCL was modest (range -0.2 to -0.8). Net effect sizes ranged from -0.15 to -0.6. Relative to the NHS real world pilot study baseline is lower in these studies (NHS baseline = 9.4 % HbA1c) and the net ES smaller (NHS ES = -1.5). In the NHS pilot study treatment with HCL brings the mean % HbA1c to 7.9 approaching a level comparable with the upper range values seen in RCTs after HCL use.

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 12 of 54

Change in HbA1c percentage: Network meta-analysis

The EAG did a frequentist random effects network meta-analysis (NMA) of the change in HbA1c percentage estimates. The NMA included 10 estimates. The reference treatment class was continuous subcutaneous insulin infusion (CSII) plus continuous glucose monitoring (CGM), where estimates more than 0 favoured CSII plus CGM. Figure 4 shows the network map for the change in HbA1c percentage over the observation period from the included studies.

Figure 4 Network map of the outcome Change in HbA1c %

==> picture [417 x 116] intentionally omitted <==

Compared with CSII plus CGM, the NMA showed that the HCL arm of the RCTs had an improvement in HbA1c percentage, that is HCL decreased the percentage HbA1c by 0.29 (95% CI: -0.37 to -0.21). The NMA results are shown in figure 5.

Figure 5 Results of the NMA of the outcome Change in HbA1c % over observation period

==> picture [417 x 105] intentionally omitted <==

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 13 of 54

Percentage time in range (between 3.9 to 10 mmol/litre)

All RCTs reported results for percentage time in range between 3.9 to 10 mmol/litre except for the study by Stewart et al. which included a pregnant population and reported time in range 3.5 to 7.8 mmol/litre. For this outcome, better glycaemic control is indicated by more time in range.

In all the RCTs the increase in percentage time in range was greater in the HCL arm than the comparator arm. The EAG said that in all cases, this was a statistically significant (p<0.05) difference. The lowest mean baseline percentage time in range was 46 to 47%, in all other studies it was over 50%. In the NHS Pilot study, baseline was 34.2% allowing considerable scope for improvement with HCL treatment which was 28.5% (unadjusted; 95% CI: 25.6 to 13.5). The change from baseline in the HCL arm of RCTs with adults of similar age range as those in the adult NHS Pilot ranged from 10% to 15%. The EAG said that the size of improvement in percentage time in range appears to be greater the lower the baseline level.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 14 of 54

Figure 6 Change from baseline in percentage time in range 3.9 to 10 mmol/litre forest plot

==> picture [417 x 409] intentionally omitted <==

----- Start of picture text -----
mean ± SD
STUDY N mean SD -2 0 2 4 6 8 10 12 14 16 18 20 AGE yr weeks BL ES
Kariyawasam HCL 17 NR NR 2 to 6 6.0 NR
7.51 (3.14,11.8)
Kariyawasam comp 17 NR NR 2 to 6 6.0 NR
Ware a HCL 34 10.10 0.18 5.6 16.0 61.5
8.7 (7.4,9.9)
Ware a comp 35 2.10 0.21 5.6 16.0 60.8
von dem Berge HCL 38 10.40 0.57 2 to 17 8.0 60.4
10.5 (8.09,12.91)
von dem Berge comp 38 -0.10 1.04 2 to 17 8.0 60.4
Collyns HCL 19 NR NR 7 to 13 4.0 NR
11.8 (8.5,15.1)
Collyns comp 19 NR NR 7 to 13 4.0 NR
Thabit HCL 32 NR NR 12 (±3.4) 12.0 NR
8.9 (5.9,11.8)
Thabit comp 33 NR NR 12 (±3.4) 12.0 NR
Ware b HCL 65 7.00 2.70 13.1 (±2.6) 26.0 47.0
6.7 (2.2,11.3)
Ware b comp 68 1.00 0.90 13.1( ±2.6) 26.0 46.0
Collyns HCL 14 NR NR 14 to 21 4.0 NR
14.4 (10.0,18.8)
Collyns comp 14 NR NR 14 to 21 4.0 NR
Tauschmann HCL 46 13.00 7.40 13 to 26 12.0 52.0
10.8 (8.2,13.5)
Tauschmann comp 40 2.00 7.90 11 to 36 12.0 52.0
Stewart HCL 16 NR NR 32 (±5) 4.0 NR
2.1 (-4.1,8.3)
Stewart comp 16 NR NR 32 (±5) 4.0 NR
Thabit HCL 25 NR NR 40 (±9.4) 12.0 NR
11.0 (8.1,13.8)
Thabit comp 24 NR NR 40 (±9.4) 12.0 NR
Benhamou HCL 63 NR NR 48.2 (±11.7) 12.0 NR
9.2 (6.4,11.9)
Benhamou comp 63 NR NR 48.2 (±11.7) 12.0 NR
Boughton HCL 20 11.30 3.60 67 16.0 69.6
8.6 (6.3,11.0)
Boughton comp 17 1.10 4.60 67 16.0 70.3
McAuley HCL 30 NR NR 67.0 16.0 NR
6.2 (8.4,8.0)
McAuley comp 30 NR NR 67.0 16.0 NR
Collyns HCL 59 NR NR 7 to 80 4.0 NR
12.5 (8.0,17.0)
Collyns comp 59 NR NR 7 to 80 4.0 NR
----- End of picture text -----

Weeks = treatment period; BL = baseline value ; comp = comparator; HCL = hybrid closed loop; N = number of participants; yr = years; ES = net effect size mean difference 95% CI [HCL vs. comparator]; medians have no error bars. NB. The population in Stewart et al., was pregnant women and the time in range refers to 3.5 to 7.8 mmol/litre rather than 3.9 to 10 mmol/litre. Collyns et

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 15 of 54

al. has 3 entries corresponding to 2 age groups (7 to 13 and 14 to 21) and all age groups combined (7 to 80).

Percentage time in range (between 3.9 to 10 mmol/litre) NMA

The EAG did a frequentist random effects NMA of the percentage time in range between 3.9 to 10 mmol/litre. The NMA included 12as shown in the network map (figure 7).

Figure 7 Network map of time in target range 3.9 to 10 mmol/litre

==> picture [417 x 116] intentionally omitted <==

The reference treatment class was CSII plus CGM, where estimates of less than 0 favoured CSII plus CGM. Compared with the CSII plus CGM treatment classification, HCL significantly increased the percentage time in range (between 3.9 to 10.0 mmol/litre), with a mean difference (MD) of 8.62 (7.03 to 10.22). The forest plot of the NMA is shown in figure 8.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 16 of 54

Figure 8 NMA results for time in target range between 3.9 to 10 mmol/litre

==> picture [417 x 130] intentionally omitted <==

Percentage time above range (over 10 mmol/litre)

For this outcome, increased percentage time in range indicates a tendency to hyperglycaemia and poor glycaemic control. For example, a negative mean difference (intervention minus comparator) indicates that more time was spent in the range more than 10 mmol/litre in the comparator group and therefore the intervention provided better glycaemic control. In all studies HCL reduced the percentage time above range more than in the comparator arms. The EAG said that the difference between arms (net effect size) was statistically significant in all cases (p < 0.05). Figure 9 shows the change from baseline in percentage time above range (over 10.0 mmol/litre) reported in the RCTs.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 17 of 54

Figure 9 Percentage time above range (over 10 mmol/litre) forest plot

||||mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median|mean ± SD || median| |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| |STUDY|N|mean|SD
-14
-12
-10
-8
-6
-4
-2
0
2||||||||||AGE yr|weeks|BL|ES| |||||||||||||||||| |Kariyawasam HCL
Kariyawasam comp|17
17|NR
NR|NR
NR||||||||||2 to 6
2 to 6|6.0
6.0|NR
NR|-5.01 (-6.21,-3.81)| |Ware a HCL
Ware a comp|34
35|-9.30
-5.00|NR
NR||||||||||5.6
5.6|16.0
16.0|32.2
36.7|-8.5 (-9.9,-7.1)| |von dem Berge HCL
von dem Berge comp|38
38|10.40
-0.10|0.57
1.04||||||||||2 to 17
2 to 17|8.0
8.0|36.3
36.3|10.5 (8.09,12.91)| |||||||||||||||||| |||||||||||||||||| |Collyns HCL
Collyns comp|19
19|NR
NR|NR
NR||||||||||7 to 13
7 to 13|4.0
4.0|NR
NR|-11.2 (-14.8,-7.6)| |Thabit HCL
Thabit comp|32
33|NR
NR|NR
NR||||||||||12 (±3.4)
12 (±3.4)|12.0
12.0|NR
NR|8.9 (5.9,11.8)| |Ware b HCL
Ware b comp|65
68|-8.00
-1.00|2.70
2.60||||||||||13.1 (±2.6)
13.1( ±2.6)|26.0
26.0|46.0
47.0|-7 (-12.5,-1.5)| |||||||||||||||||| |||||||||||||||||| |Collyns HCL
Collyns comp|14
14|NR
NR|NR
NR||||||||||14 to 21
14 to 21|4.0
4.0|NR
NR|-14 (-18.4,-9.55)| |Tauschmann HCL
Tauschmann comp|46
40|-12.00
-2.00|2.00
2.35||||||||||13 to 26
11 to 36|12.0
12.0|44.0
44.0|-10 (-13.2,-7.5)| |||||||||||||||||| |||||||||||||||||| |Stewart HCL
Stewart comp|16
16|NR
NR|NR
NR||||||||||32 (±5)
32 (±5)|4.0
4.0|NR
NR|-0.1 (-4.2,4.0)| |Thabit HCL
Thabit comp|25
24|NR
NR|NR
NR||||||||||40 (±9.4)
40 (±9.4)|12.0
12.0|NR
NR|-9.6 (-13.0,-6.3)| |Benhamou HCL
Benhamou comp|63
63|NR
NR|NR
NR||||||||||48.2 (±11.7)
48.2 (±11.7)|12.0
12.0|NR
NR|-6.8 (-9.7,-3.9)| |Boughton HCL
Boughton comp|20
17|-8.80
-4.10|0.00
0.00||||||||||67
67|16.0
16.0|25.5
25.5|-8.5 (-10.9,-6.1)| |McAuley HCL
McAuley comp|30
30|NR
NR|NR
NR||||||||||67.0
67.0|16.0
16.0|NR
NR|-5.4 (-7.3,-3.5)| |Collyns HCL
Collyns comp|59
59|NR
NR|NR
NR||||||||||7 to 80
7 to 80|4.0
4.0|NR
NR|-12.1 (-16.8,-7.38)| |N = number of participants c
duration; BL = mean baselin
in % in range in HCL arm rel
were usually statistically adj
net ES. Ware and Boughton
have no error bars.
Percentage time above ran||||ont
e v
ati
ust
stu
ge|rib
alu
ve
ed.
di
(o|uti
e ;
to c
Be
es
ve|ng
ES
on
nh
rep
r 1|dat
=
tro
am
ort
0 m|a;
ne
l a
ou
ed
m|yr
t e
rm,
an
me
ol/|= y
ffe
n.
d
di
litr|ears; weeks = treatment
ct size comparing reduction
b. the ES values reported
Thabit and only reported
an values. Median values
e) NMA|||||

N = number of participants contributing data; yr = years; weeks = treatment duration; BL = mean baseline value ; ES = net effect size comparing reduction in % in range in HCL arm relative to control arm, n.b. the ES values reported were usually statistically adjusted. Benhamou and Thabit and only reported net ES. Ware and Boughton studies reported median values. Median values have no error bars.

Percentage time above range (over 10 mmol/litre) NMA

The NMA included the same 12 estimates as those in the time in range (between 3.9 to 10 mmol/litre) NMA (see network map in figure 7). The

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 18 of 54

reference treatment class was CSII plus CGM, where estimates over 0 favoured CSII plus CGM. Compared with CSII plus CGM, HCL significantly decreased time above range (percentage above 10.0 mmol/litre), with a mean difference (MD) of -7.2% (95% CI -8.92 to -5.48). The NHS Pilot study reported an unadjusted reduction in time above range of 14 mmol/litre or over (rather than 10 mmol/litre) of 22.2 %.

Figure 10 NMA results for time in target range (% more than 10 mmol/litre)

==> picture [417 x 110] intentionally omitted <==

Percentage time below range (less than 3.9 mmol/litre)

For this outcome, a positive mean difference (intervention minus comparator) indicates that more time was spent in the range less than 3.9 mmol/litre in the intervention group and so there was a higher risk of hypoglycaemia for the intervention group compared with the comparator. The EAG said that because of skewed data, results were mostly reported as medians with IQRs, with only a few studies reporting mean (plus or minus sd). Figure 11 shows the percentage time below range (less than 3.9 mmol/litre) reported in the RCTs. The mean or median percentage time below range at baseline was small (6% or less), the ES was also small occasionally reaching statistical significance. The NHS Pilot study did not report this outcome.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 19 of 54

Figure 11 Percentage time below range less than 3.9 mmol/litre forest plot

==> picture [417 x 296] intentionally omitted <==

----- Start of picture text -----
mean or
mean ± SD || median median
STUDY N mean SD -5 -4 -3 -2 -1 0 1 2 3 4 5 AGE yr weeks BL ES
Kariyawasam HCL 17 NR NR 2 to 6 6.0 NR
-2.62 (-4.22,-1.01)
Kariyawasam comp 17 NR NR 2 to 6 6.0 NR
Ware a HCL 34 -0.70 0.16 5.6 16.0 4.50
0.1 (-0.4,0.5)
Ware a comp 35 -0.40 0.16 5.6 16.0 3.90
Collyns HCL 19 -0.20 0.26 7 to 13 12.0 NR
10.5 (8.09,12.91)
Collyns comp 19 0.10 0.17 7 to 13 12.0 NR
Ware b HCL 65 NR NR 13.1 (±2.6) 26.0 6.10
-0.53 (-1.78,2.83)
Ware b comp 68 NR NR 13.1( ±2.6) 26.0 5.40
Collyns HCL 14 NR NR 14 to 21 26.0 NR
8.9 (5.9,11.8)
Collyns comp 14 NR NR 14 to 21 26.0 NR
Tauschmann HCL 46 -0.90 0.00 13 to 26 12.0 3.50
-0.83 (-1.4,-0.16)
Tauschmann comp 40 0.60 0.00 11 to 36 12.0 3.30
Benhamou HCL 63 NR NR 48.2 (±11.7) 12.0 NR
-2.4 (-3.0,-1.7)
Benhamou comp 63 NR NR 48.2 (±11.7) 12.0 NR
Boughton HCL 20 -0.10 0.00 67 26.0 1.80
-0.1 (-0.3,0.2)
Boughton comp 17 0.10 0.00 67 26.0 1.60
McAuley HCL 30 NR NR 67.0 12.0 1.21
-0.47 (-1.05,-0.25)
McAuley comp 30 NR NR 67.0 12.0 1.69
Collyns HCL 59 NR NR 7 to 80 16.0 NR
-0.4 (-1.1,0.28)
Collyns comp 59 NR NR 7 to 80 16.0 NR
----- End of picture text -----

N = number of participants contributing data; yr = years; weeks = treatment duration; BL = mean baseline value ; ES = net effect size comparing reduction in % in range in HCL arm relative to control arm, n.b. the ES values reported were usually statistically adjusted

Some of the RCTs also reported percentage time below range less than 3.0 mmol/litre. The mean or median percentage time below range was less than 1.5% in both arms (see table 3, in the external assessment report) and ES values (HCL compared with comparator) reported were very small. These are shown in figure 12. This outcome was reported in the NHS Pilot study. The percentage times below range were reported as: baseline 0.36%; follow up 0.34%; providing a difference for HCL of -0.02 (95%CI : -0.01 to 0.2).

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 20 of 54

Figure 12 Percentage time below range less than 3.0 mmol/litre forest plot

==> picture [417 x 244] intentionally omitted <==

----- Start of picture text -----
mean or
mean ± SD || median median
STUDY N mean SD -0.40 -0.20 0.00 0.20 0.40 AGE yr weeks BL ES
Kariyawasam HCL 17 NR NR 2 to 6 6.0 NR
-0.44 (-0.96,0.08)
Kariyawasam comp 17 NR NR 2 to 6 6.0 NR
Ware a HCL 34 -0.70 0.16 5.6 16.0 0.80
0.02 (-0.1,01)
Ware a comp 35 -0.40 0.16 5.6 16.0 0.60
von dem Berge HCL 38 -0.20 0.26 7 to 13 12.0 0.80
0.2 (0.04,0.36)
von dem Berge comp 38 0.10 0.17 7 to 13 12.0 0.80
Collyns HCL 19 NR NR 13.1 (±2.6) 26.0 NR
-0.2 (-.42,0.02)
Collyns comp 19 NR NR 13.1( ±2.6) 26.0 NR
Collyns HCL 14 NR NR 14 to 21 26.0 NR
-0.01 (-0.26,0.06)
Collyns comp 14 NR NR 14 to 21 26.0 NR
Boughton HCL 20 NR NR 13 to 26 12.0 NR
0.0 (-0.1,0.1)
Boughton comp 17 NR NR 11 to 36 12.0 NR
McAulery HCL 30 NR NR 48.2 (±11.7) 12.0 NR
-0.11 (-0.16,-0.05)
McAuley comp 30 NR NR 48.2 (±11.7) 12.0 NR
Collyns HCL 59 5.00 NR 67 26.0 NR
-0.1 (-0.31,0.11)
Collyns comp 59 5.00 NR 67 26.0 NR
----- End of picture text -----

N = number of participants contributing data; yr = years; weeks = treatment duration; BL = mean baseline value ; ES = net effect size comparing reduction in % in range in HCL arm relative to control arm, n.b. the ES values reported were usually statistically adjusted

Percentage time below range (less than 3.9 mmol/litre) NMA

The NMA included 7 estimates from 7 studies as shown in the network map (figure 13). The reference treatment class was CSII plus CGM, where estimates more than 0 favoured CSII plus CGM.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 21 of 54

Figure 13 Network map of time below range less than 3.9 mmol/litre

==> picture [417 x 116] intentionally omitted <==

Although there was a mean difference of less than 0 (that is, favouring HCL) The EAG said there was no statistically significant difference between HCL and CSII plus CGM. The NMA forest plot is shown in figure 14.

Figure 14 NMA results for time below range less than 3.9 mmol/litre

==> picture [417 x 101] intentionally omitted <==

Subgroup analyses

The EAG did a subgroup analysis where studies were categorised based on mean or median age of participants at baseline. Participants less than 18 years were classed as children and young adults, participants 18 years and over were classed as adults. The NMA results in the subgroups were similar to those in the whole population. The change in HbA1c percentage for HCL was -0.31 (-0.43, -0.20) in the children and young adults subgroup and -0.24 (-0.32, -0.15) in the adult subgroup. Find full details and results of the subgroup analyses on pages 104 to 105 of the external assessment report.

Intermediate outcomes: Observational studies

The EAG said that the outcome estimates reported for observational studies were quantitatively broadly in line with those from the RCTs. Measures of

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 22 of 54

glycaemic performance such as HbA1c percentage, percentage time in range (3.9 to 10 mmol/litre), and percentage time above range (over 10mmol/litre) all improved on transfer to HCL (or to AHCL) without any strong evidence that hypoglycaemia became more of a problem. However, the EAG said that changes in hypoglycaemia were mostly underpowered in these studies.

Change in HbA1c percentage

HbA1c percentage improved on transfer to HCL (or to an advanced HCL). The range of change was narrow across RCTs and single arm studies. The improvement in HbA1c percentage level was much greater in the NHSE adult pilot study, however the EAG said that in this study, the baseline level was considerably above that in all other studies (around 9.4%) and so there was greater scope for improvement. In the NHS Pilot with children and young people (CYP) baseline HbA1c was lower (around 7.8%) and benefit more modest (-0.70%). Figure 15 shows the change in HbA1c percentage from baseline in study participants receiving HCL intervention.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 23 of 54

Figure 15 Change in HbA1c % from baseline in study participants receiving HCL intervention

==> picture [417 x 318] intentionally omitted <==

Percentage time in range (between 3.9 to 10 mmol/litre)

Most studies had a baseline time in range (3.9 to 10 mmol/litre) above 50%. In the NHSE adult pilot adult study, the baseline time in range was 34.2%. The EAG said that this likely reflects the broad inclusion of patients and indicates that along with the higher HbA1c baseline, that people in this study had poor glycaemic control before receiving the HCL intervention. Similarly, in the NHSE CYP pilot study, the baseline time in range was relatively poor at 48.7%. In the NHSE adult pilot, the benefit from HCL was larger (28.5%) than the other studies with a mean value at the end of follow up of 62.7%. The EAG said that this end of follow up value was similar to the values from the other observational studies. In the NHSE CYP pilot, the end of study time in range was also similar at 63%. Figure 16 shows a forest plot of percentage

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 24 of 54

time in range (between 3.9 and 10 mmol/litre) in study participants from both RCTs and observational studies that had HCL intervention.

Figure 16 Change from baseline of percentage time in range (3.9 to 10 mmol/litre)

==> picture [417 x 349] intentionally omitted <==

----- Start of picture text -----
median
or mean ± SD or median
STUDY N mean SD 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 AGE yr weeks BL
Ware a HCL 34 10.10 1.90 5.6 16 61.5
von dem Berge HCL 38 10.40 2.30 2 to 17 8 60.4
Breton HCL 77 14.00 2.30 11.0 16 53
Ware b HCL 65 7.00 2.70 12.0 12 47
Abraham HCL 67 9.40 2.20 13.1 26 53.1
Tauschmann HCL 46 13.00 7.40 15.0 26 52
Brown HCL 112 13.00 2.00 22.0 12 61
McAuley HCL 30 6.20 8.00 67.0 26 NR
Boughton HCL 20 10.30 3.60 67.0 16 69.6
NHS Pilot HCL 456 28.50 1.50 40.0 26 34.2
Forlenza HCL 46 8.10 4.30 48.0 12 55.7
Bergenstahl HCL 113 6.00 1.00 68.0 16 57
Bergenstahl AHCL 113 10.00 1.00 26.0 26 57
Bassi AHCL all 90 14.60 1.70 14 to29 12 NR
Beato-Vibora AHCL 52 12.80 2.20 14 to 29 12 67.1
Breton HCL 7801 10.30 0.15 43.0 12 63.2
Carlson AHCL 39 10.30 1.82 14 to 21 12 62.4
Carlson AHCL 118 4.20 1.13 22 to 75 12 70.9
NHS Pilot CYP 251 14.30 1.10 44.2 26.0 48.7
----- End of picture text -----

Percentage time above range (over 10 mmol/litre)

All studies reported an improvement from baseline, ranging from 3.0% to 14% reduction in percentage time above range. The NHSE adult pilot study did not report this outcome but did report unadjusted (uncorrected) percentage time above range (above 14 mmol/litre). At baseline the percentage time above 14 mmol/litre was 37.4% and a further 26.6% of time was in the range between 10 and 14 mmol/litre, indicating that at baseline the NHSE Pilot study participants had a large percentage of time in the hyperglycaemic state

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 25 of 54

(around 64% of time). Transfer to HCL resulted in large reduction of 22.6% time above the 14 mmol/litre range. The benefit of HCL in the range 10 to 14 mmol/litre was more modest (a reduction in time above range of 4%). The EAG said that these results suggest that HCL improved hyperglycaemia considerably in the upper range but that a substantial proportion remained slightly above the 10 mmol/litre cut off. Figure 17 shows a forest plot of the change from baseline in the percentage time above range (above 10 mmol/litre).

Figure 17 Change from baseline of percentage time above range (above 10 mmol/litre)

==> picture [417 x 346] intentionally omitted <==

----- Start of picture text -----
median
mean ± SD or median
or
STUDY N mean -16.0 -12.0 -8.0 -4.0 0.0 AGE yr weeks BL
Ware a HCL 34 -9.30 5.6 16 32.2
von dem Berge HCL 38 -10.50 2 to 17 8 36.3
Breton HCL 77 -14.00 11 16 45
Ware b HCL 65 -8.00 13.1 26 46
Abraham HCL 67 -7.40 15 26 41.8
Tauschmann HCL 46 -12.00 22 12 44
Brown HCL 112 -9.00 33 26 36
McAuley HCL 30 -5.40 67.0 16 NR
Boughton HCL 20 -4.00 68 16 25.5
Forlenza HCL 46 -8.00 2 to 7 12 41
Bergenstahl HCL 113 -7.00 14 to 29 12 41
Bergenstahl AHCL 113 -10.00 14 to 29 12 41
Bassi AHCL all 90 -5.70 24.4 4 NR
Beato-Vibora AHCL 52 -12.60 43 12 29.4
Breton HCL 7801 -5.50 6 to 91 52 25.2
Carlson AHCL 39 -9.40 14 to 21 12 34.3
Carlson AHCL 118 -3.10 22 to 75 12 25.7
Carlson AHCL 118 -3.10 22 to 75 12 8.3
----- End of picture text -----

Percentage time below range (less than 3.9 mmol/litre)

The change in percentage time below range (less than 3.9 mmol/litre and less than 3.0 mmol/litre) was reported in most observational studies. Both

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 26 of 54

percentage time below 3.9 mmol/litre at baseline (range from 2.1% in the NHS Pilot adult study to 3.4%) and after HCL intervention were small, with a resulting mean improvement of around 1% or less. The NHS pilot adult study reported a change of -0.5% and an associated p value of less than 0.001. The NHSE CYP pilot study also reported a statistically significant improvement. Only 1 other study (Carlson et al., adult patients) reported a statistically significant improvement (p less than 0.05).

Figure 18 shows the mean (95% CI) change from baseline in percentage time below 3.9 mmol/litre; confidence intervals were wide. Some of the single arm studies reported other outcomes indicative of hypoglycaemic status, most commonly percentage time below range less than 3.0 mmol/litre. The results are shown in figure 19.

Figure 18 Mean (95% CI) change from baseline in percentage time below range less than 3.9 mmol/litre

==> picture [417 x 222] intentionally omitted <==

----- Start of picture text -----
mean 95% CI
STUDY N mean -3.0 -2.0 -1.0 0.0 1.0 AGE yr weeks BL
NHS Pilot adult 540 -0.50 24.4 26 2.1
Forlenza 46 -0.10 2 to 7 12 3.3
Bergenstahl HCL 113 -0.20 14 to 29 12 2.3
Bergenstahl AHCL 113 -0.20 14 to 29 12 2.3
Bassi all 90 -0.30 24.4 4 NR
Beato-Vibora 52 -1.10 43.0 12 3.4
Carlson 39 -1.00 14 to 21 12 3.3
Carlson 118 -1.00 22 to 75 12 3.4
NHS Pilot CYP 20 -1.20 2 to 19 26 3.6
----- End of picture text -----

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 27 of 54

Figure 19 Mean (95% CI) change from baseline in percentage time below range less than 3.0 mmol/L

==> picture [421 x 157] intentionally omitted <==

----- Start of picture text -----
mean 95% CI
STUDY N mean -3.0 -2.0 -1.0 0.0 1.0 AGE yr weeks BL
NHS Pilot 540 -0.02 24.4 26 0.4
Forlenza 46 0.00 2 to 7 12 0.7
Bassi all 90 -0.27 24.4 4 NR
Beato-Vibora 52 -0.20 43.0 12 0.9
Carlson 118 -0.30 14 to 22 12 0.8
Carlson 39 -0.30 23 to 75 12 0.9
----- End of picture text -----

Clinical outcomes

The EAG said that the studies did not consistently report any additional outcomes, however some of the RCTs did report on adverse events.

Adverse events

The EAG said that the RCTs reported a low number of adverse events for both treatment groups. Although some reports of hypoglycaemia were identified in the included studies, the EAG did not identify any clear trends and differences between HCL and the comparator. In the study by Benhamou et al. 2022, one severe hypoglycaemia event and one ketoacidosis event were reported in 2 different participants. The ketoacidosis occurred while the patient was under closed loop. The severe hypoglycaemia occurred while the patient had temporarily switched to open loop treatment.

Patient reported outcomes

The EAG said that there were several studies that used various tools and different survey approaches to report technology satisfaction. Only 1 study (Benhamou et al. 2022), comparing an open loop to a closed loop system, found that user satisfaction had increased significantly after the closed loop period. Other studies did not observe any significant changes.

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 28 of 54

The FLAIR study, reported mean scores on the glucose monitoring satisfaction survey at screening, at the end of the period using the HCL system and at the end of the period using the advanced HCL system. Emotional burden and behavioural burden satisfaction subscales were significantly improved with the advanced HCL system.

The study by Tauschmann et al. 2018 used the Paediatric Quality of Life Inventory (PedsQL) questionnaire. This was given to participants and guardians of participants under 17 years, before and after the intervention period. The use of the closed-loop system was not associated with any additional burden.

McAuley et al. 2022 used the hypoglycaemia fear survey score and reported no significant difference between HCL and sensor augmented pump (SAP) groups.

The study by Wheeler et al. 2022 reported patient reported outcomes from the Collyns et al. 2021 study. It compared technology satisfaction and sleep quality between advanced HCL and SAP plus predictive low-glucose management (PLGM). Overall treatment satisfaction was significantly higher for the advanced HCL group compared to the SAP plus PLGM group. There was no significant difference in anticipated worry of hypoglycaemia. Results showed no changes in the well-being index and hypoglycaemia fear or confidence.

External submissions

Medtronic submission

The Medtronic clinical effectiveness submission compared the (Advanced) HCL systems with real time CGM plus continuous subcutaneous insulin infusion (non-integrated). Find full details of the submission, including a critique by the EAG, on pages 115 to 119 in the external assessment report.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 29 of 54

Dexcom submission

The Dexcom clinical effectiveness submission compared HCL with sensor augmented pump (SAP) systems, based on the results of 1 systematic review and NMA, and 8 RCTs. Find full details of the submission, including a critique by the EAG, on pages 119 to 123 in the external assessment report.

CamDiab submission

The CamDiab clinical effectiveness submission included 10 studies. Find full details of the submission, including a critique by the EAG, on pages 123 to 127 in the external assessment report.

Tandem submission

The Tandem clinical effectiveness submission included a poster and 2 papers (1 unpublished). Find full details of the submission, including a critique by the EAG, on pages 127 to 129 in the external assessment report.

3 Cost effectiveness evidence

The EAG did a systematic review to identify any published economic evaluations of hybrid closed loop systems for managing blood glucose levels in people with type 1 diabetes. Find the full systematic review results on pages 134 to 146 of the external assessment report. The EAG also constructed a de novo economic model to assess the cost effectiveness of hybrid closed loop systems for managing blood glucose levels in people with type 1 diabetes.

Systematic review of cost-effectiveness evidence

The EAG identified 6 studies that were included in the review, 5 were economic evaluations of hybrid closed loop (HCL) systems and 1 was a budget impact analysis. Four of the economic evaluation studies (Jendle et al. 2019, Jendle et al. 2021, Roze et al. 2021, and Serne et al. 2022) used the IQVIA CORE Diabetes Model (CDM). One study by the Scottish Health

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 30 of 54

Technologies Group (SHTG) used the Sheffield type 1 diabetes model (Harbour et al. 2022). The budget impact analysis was done by the Canadian Agency for Drugs and Technology in Health (CADTH) and used a customised Microsoft Excel tool.

The economic evaluation studies compared the cost effectiveness of HCL systems with various diabetes management technologies (for example, isCGM plus MDI, CSII and self-monitoring of blood glucose). Two of the 6 studies were done in Sweden (Jendle et al. 2019 and Jendle et al. 2021), and 1 each in the UK (Roze et al. 2021), Netherlands (Serne et al. 2022), Scotland (Harbour et al. 2022) and Canada (CADTH, 2021). The studies were assessed using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) and Phillips checklists where applicable.

The EAG said that there was substantial heterogeneity in the choice of baseline cohort data and treatment effects data. Only the SHTG study used baseline data for its population of interest.

The EAG said that structure of the models used in the cost effectiveness studies were good quality. The IQVIA CDM and the Sheffield type 1 diabetes model are validated models for evaluating diabetes technologies. It also said that the IQVIA CDM is capable of capturing both long and short term clinical complications and costs associated with T1DM and has been extensively validated for use in this condition.

The EAG said that in 4 of the cost effectiveness studies, the base case results were very sensitive to the severe hypoglycaemic rates (SHE) and changes in the assumptions relating to the quality-of-life benefit associated with reduced fear of hypoglycaemia (FOH). It also said that the cost effectiveness acceptability curves from these studies showed that HCL systems are expected to be cost effective compared with the comparator technologies at various hypothetical maximum acceptable thresholds.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 31 of 54

Company cost effectiveness submissions

Medtronic submission

The Medtronic submission used the IQVIA CDM to compare the advanced HCL 780G Minimed pump with CSII using the 640G Minimed pump. Find full details of the submission and the EAG’s observations on pages 154 to 156 of the external assessment report.

Dexcom submission

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. Find full details of the

submission and the EAG’s observations on pages 157 to 159 of the external assessment report.

CamDiab submission

The CamDiab submission presented two cost effectiveness modelling exercises, one based upon the Dan05 study among patients aged 6 to 18 years using the XXXXXXXXXXXXXXX and the other based upon the KidsAP02 study among patients aged 1 to 7 years using XXXXXXXXX Find full details of the submission and the EAG’s observations on pages 159 to 163 of the external assessment report.

Tandem submission

The Tandem submission referenced the Dexcom submission economics and provided no additional cost effectiveness estimates.

Economic analysis

The EAG said that the IQVIA CDM and the Sheffield type 1 diabetes model are both suited to economic analyses of diabetes management technologies allowing for deterministic and probabilistic sensitivity analyses to be done. The IQVIA CDM uses time, time in state and diabetes dependent probabilities to simulate progression of diabetes and diabetes related complications with both diabetes and non-diabetes mortality accounted for. It allows clinical and cost

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 32 of 54

data to be inputted directly into the model or for default parameters to be used. The EAG preferred the IQVIA CDM. It used real world data from the UK as a simulation cohort. Further details on clinical and cost inputs are presented in the following sections.

Model structure

The modelled treatment pathway assumes that people remain on a single treatment option throughout: either CSII plus CGM (intermittently scanned or real time), PLGS or HCL.

In line with DG21 and NG17 the EAG used the IQVIA CDM to model the micro and macro vascular complications of diabetes and patients’ overall survival. The IQVIA CDM predicts the progress of people with T1DM over their lifetime, modelling the incidences of the 11 macro and micro vascular complications, the likelihoods of which are affected by T1DM. Figure 20 shows the structure of the IQVIA CDM.

Figure 20 IQVIA CDM structure

==> picture [417 x 171] intentionally omitted <==

The EAG said that the default and recommended setting is to sample 1,000 people from the patient characteristics and run each of these people through the model 1,000 times. The IQVIA team advised the EAG that for modelling a T1DM cohort only the non-specific mortality approach should be used. The IQVIA CDM models deaths from myocardial infarction, congestive heart

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 33 of 54

failure, stroke and renal disease. Therefore, the EAG removed deaths due to cardiovascular disease, cerebrovascular disease and renal failure from the England and Wales life table (2015 to 2017) to determine non-specific mortality estimates. Deaths due to hypertension were also removed in a scenario analysis.

The EAG said it had concerns about the reliability of using the IQVIA CDM to model a paediatric population due to key sources using data that relates to an adult population. It said that the model is affected by both the longer duration that is required for a lifetime horizon and the degree to which the risk equations of the model relate to a paediatric population. The EAG did an exploratory analysis using the NMA results for the subset of paediatric studies and a scenario analysis that applies the NHSE paediatric pilot results (see exploratory paediatric modelling results section). This analysis is shown in appendix 5 of the external assessment report.

Modelling of other clinical effects

The EAG said that there was a lack of clarity around the IQVIA CDM implementation of the quality of life decrements for non severe hypoglycaemic events (NSHEs). The EAG used the IQVIA CDM to model the effects of HbA1c on survival and the micro and macro vascular complications of diabetes. The IQVIA CDM overall survival curve for each technology is then coupled with technology specific treatment costs and comparator specific NSHE and severe hypoglycaemic event (SHE) rates (in scenario analyses). With the addition of the events’ unit costs and disutilities this enables technologies’ other effects to be incorporated into the cost effectiveness analysis.

Perspective, discount rates and time horizon

The perspective for costs is the NHS and PSS, the perspective for benefits is that of the patient, and costs and benefits are discounted at 3.5%. The base case assumes a 50 year time horizon which the EAG said is effectively a

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 34 of 54

lifetime horizon for most patients. Because of the uncertainty around the IQVIA CDM outputs for longer time horizons, the EAG did scenario analyses which explored time horizons of 8, 12 and 24 years.

Population

The EAG used data from the 2019 to 2020 National Diabetes Audit subgroup of those on pump therapy for the key baseline characteristics. In a scenario analysis it also used data from the NHSE adult pilot study. Table 2 shows the population baseline characteristics.

Table 2 Population baseline characteristics

For other baseline characteristics needed as inputs to the IQVIA CDM, the EAG took them from NG17, which uses data from the Repose trial comparing pumps with multiple daily injections. These characteristics relate to a slightly more severe controlled group of people with a baseline HbA1c of 9.1%. Full details of these additional baseline characteristics are in table 39 (appendix 7) of the external assessment report.

Comparators

In addition to the intervention (HCL), the cost effectiveness analysis considered the 2 comparators in the EAG network meta-analysis (NMA):

• CSII plus CGM non-integrated

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 35 of 54

• LGS/PLGS

The EAG did not evaluate CSII plus CGM separately as CSII plus real time CGM (rtCGM) and CSII plus intermittently scanned CGM (isCGM). It assumed the balance to be 10% CSII plus rtCGM and 90% CSII plus isCGM for adult patients. However, the EAG said that this may underestimate CSII plus isCGM use. In the scenario analysis that uses the NHSE adult pilot data, the EAG assumed that CSII plus CGM was 100% CSII plus isCGM due to prior use of CSII plus isCGM being reported as a requirement. Because LGS/PLGS systems are no longer available to purchase in the UK, this may not be a relevant comparator.

Model inputs

Modelling of HbA1c effects: HbA1c progression

In the base case analysis the EAG assumed no annual worsening of HbA1c over time. However, as the IQVIA CDM default for HbA1c progression applies an annual worsening of 0.045%, the EAG included this as a scenario analysis, applied to both the intervention and comparator arms.

HbA1c effects

In the base case, the EAG used the results of the RCT NMA (see section 2, intermediate outcomes -RCTs). Two scenario analyses were done. One that restricted the NMA evidence base to adult trials, and 1 where the mean HbA1c percentage change of the NHSE adult pilot was used (applied to the NHSE adult pilot baseline characteristics). The base case assumes that the HbA1c effect endures for the model time horizon of 50 years. Scenario analyses that use durations of 5 years, 10 years and 20 years were also done. Table 3 shows the mean HbA1c percentage changes for each technology.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 36 of 54

Table 3 HbA1c percentage changes

Non severe hypoglycaemic event (NSHE) and severe hypoglycaemic event (SHE) rates

NSHE rates were not reported in the RCTs. The EAG did not include NSHE or SHE effects in its base case. The EAG did scenario analyses that estimate NSHE and SHE rates based upon estimates in the literature coupled to the EAG NMA results for time below range.

For NSHEs the EAG did a scenario analysis that couples the annual NSHE rate for HCL of 20.8 (Brown et al. 2019 and Breton et al. 2022) with the EAG NMA time below 3.0 mmol/litre net effect estimates, the weighted mean of the end of trials’ time below 3.0 mmol/litre for the CSII plus CGM and the assumption that the number of NHSEs is proportionate to the time below 3.0 mmol/litre. The EAG estimates of NSHEs and SHEs used in the main scenario analysis are shown in table 4.

Table 4 EAG estimates of NHSEs and SHEs for main scenario analysis

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 37 of 54

Scenarios of annual NSHE rates of 57.2 (from Abraham et al. 2021) and 13.0 (from Kariyawasam et al. 2021) for HCL are also presented. For SHEs, the EAG used the same approach in exploratory scenarios that assume SHE rates are proportionate to time below 3.0 mmol/litre, coupled with the annual SHE rate for HCL of 0.26 (reported in McAuley et al. 2020). Find details of NSHE and SHE rates on pages 169 to 175 of the external assessment report.

Inputs for exploratory paediatric modelling

In the exploratory analysis the EAG revised the key baseline characteristics to reflect the NHSE paediatric pilot baseline data. These are shown in Table 5.

Table 5 Exploratory paediatric modelling: baseline characteristics

The base case used the NMA HbA1c results for the subset of paediatric studies and a scenario analysis was done that used the HbA1c results from the NHSE paediatric pilot study. The HbA1c model inputs are shown in table 6.

Table 6 Exploratory paediatric modelling: HbA1c (s.e.) changes

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 38 of 54

Because of the lower mean baseline age, the EAG extended the time horizon to 80 years. The EAG also assumed that paediatric patients had not developed any of the complications associated with diabetes and modelled by the IQVIA CDM.

Costs

Training costs

The EAGs base case does not include training costs involved from moving from MDI plus CGM to CSII plus CGM or to HCL, because estimates for these in terms of staff time and outpatient visits were the same. However, moving from CSII plus CGM to HCL, with most patients moving from isCGM to rtCGM results in a training cost of £1,132.

Treatment costs

The EAG used current list prices for the technologies provided by the NHS Supply Chain. It said that the costs of HCL pumps and consumables differ slightly between systems but the total 4 year costs are similar, except for 1 system which is around an annual average of £500 more than the unweighted average. This also applies to the LGS/PLGS systems.

The EAG used the unweighted averages for year 1 and years 2, 3 and 4 (see table 7) and provides a scenario analysis which increases these by £500 for both HCL and LGS/PLGS. To account for potential reductions in CGM sensor durations, the EAG increased the cost of all CGM sensors by 5%.

Table 7 Pump and consumable costs

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 39 of 54

For insulin costs, the EAG added an additional annual average of £315 to all regimes based on a daily average of 50 IU.

Companies indicated that prices will change for the next financial year and some products have confidential volume discounts. The EAG has addressed these aspects in the separate confidential appendix.

Ongoing visits and costs of micro and macro vascular complications

The EAG assumed that that without complications the average patient once established on treatment is seen in an outpatient clinic once per quarter, at an annual routine outpatient cost of £640. Other ongoing routine management costs and costs of micro and macro vascular complications are taken from NG17 and inflated to 2019 to 2020 prices. Find details of these costs in tables 27 and 28 (pages 200 to 201) in the external assessment report.

NSHE and SHE costs

Where NSHEs and SHEs are included in scenario analyses, the EAG applied a cost of £1.83 for SHEs not requiring outside medical attention and of £542 for those requiring medical attention. It assumed that 37.9% of SHEs require medical attention. The EAG also did a scenario analysis that increased the cost of SHEs not requiring outside medical attention to £36 and those requiring medical attention to £628. Another scenario analysis costs all SHEs at the 2021 updated cost of £381 of NG17.

Health-related quality of life and QALY decrements

The EAG used a value of 0.839 for quality of life without complications for patients with T1DM. This was based on the EQ-5D baseline average reported by Peasgood et al. 2016.

Disutilities of micro and macro vascular complications

Disutilities of micro and macro vascular complications are taken from the default values of the IQVIA CDM, in line with NG17. Find details of these in table 23, page 187 of the external assessment report.

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 40 of 54

Disutilities of hypoglycaemia events

The EAG said that for the disutility of NSHEs (used only in scenario analyses), the studies by Gordon et al. and Currie et al. provide estimates that conform most closely to the NICE reference case.

Hypoglycaemia events and carer disutilities

The EAG did not identify any data that quantified disutilities associated with impact of hypoglycaemic events on parents and carers. It said that a reasonable upper limit for the effect upon carers might be to assume that they have the same disutility as the person with T1DM that they are caring for. The EAG did a scenario analysis that doubles the disutilities associated with hypoglycaemia events to reflect possible effects on carers.

Base case results

The base case modelling disaggregate results are shown in table 29 of the external assessment report (see page 203). These results showed that compared with CSII plus CGM, the use of HCL is estimated to increase undiscounted survival by 0.458 years. Discounting reduces the net survival gain to 0.149 years, giving a patient gain of 0.160 QALYs. The net treatment cost of £31,185 is partly offset by renal savings of £421 and eye savings of £3,085, resulting in a net cost of £28,628.

Cost effectiveness of HCL

The base case results suggest that PLGS is extendedly dominated by HCL, and that HCL has a cost effectiveness estimate of £179k per QALY gained. The EAG’s base case cost effectiveness estimates are shown in table 8.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 41 of 54

Table 8 Base case cost effectiveness estimates

The EAG said that the IQVIA CDM does not allow periodic capital costs to be modelled, so for the deterministic modelling it used the modelled OS curves to estimate treatment costs.

Analysis of alternative scenarios

Find the full list of the EAG’s scenario analyses on pages 205 to 206 of the external assessment report.

Scenario analyses

In the scenario analyses, the EAG said that PLGS was extendedly dominated throughout and therefore was not shown in the results. Table 9 shows the ICERs for HCL compared with CSII plus CGM.

Table 9 Scenario analyses’ ICERs: HCL vs CSII+CGM

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 42 of 54

The scenarios with the largest effect on the ICERs were when the NHS adult pilot baseline characteristics and HbA1c effects were used, either with or without reducing the modelled complication costs (SA02b and SA02c). These scenarios reduced the ICERs to £12,398 per QALY and £21,583 per QALY, respectively. All other scenarios that used baseline characteristics from the 2019 to 2020 National Diabetes Audit and HbA1c effects from the NMA had ICERs above £100k, ranging from £109k to £910k per QALY.

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 43 of 54

Threshold price analyses

The EAG did a threshold price analysis around the average annual cost of HCL that would result in ICERs of £20,000 and £30,000 per QALY. This is shown in the confidential appendix.

Exploratory paediatric modelling

The exploratory paediatric modelling base case disaggregate results are shown in appendix 5 table 34 in the external assessment report. A summary of the exploratory paediatric modelling base case results is shown in table 10.

Table 10 Exploratory paediatric modelling: base case cost effectiveness estimates

As for the adult modelling, PLGS was extendedly dominated by HCL. Compared with CSII plus CGM, the use of HCL is estimated to increase undiscounted survival by 0.819 years. The additional treatment costs of £40,606 are partially offset by savings in renal complications of £2,459 and eye diseases of £5,143 resulting in total net costs of £32,966. With the gain of 0.196 QALYs this gives an ICER of £168,196 per QALY gained.

The EAG did a range of scenario analyses, with resulting ICERs ranging from £25,868 to £191k per QALY. Including only paediatric RCT studies reduced the ICER to £116k per QALY. Using the NHSE CYP pilot HbA1c change of - 0.7% improved the ICER to £54,727 per QALY. Including the quality-of-life effects of the improvements reported in the hypoglycaemia fear survey (HFS2-ws) during the pilot further improves the cost effectiveness to £35,259 per QALY. If both parents also have a similar quality of life improvement for 15

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 44 of 54

years it improves to £25,868 per QALY. Reducing the cost of complications to account for their possible overestimation worsens the cost effectiveness to £69,013 per QALY. Full details of the scenario analyses and results are in appendix 5 table 36 in the external assessment report. The EAG said that in all the scenario analyses the HbA1c effect, the HFS2-ws effect and the composition of CSII+CGM may change as the patient moves from childhood into adulthood

4 Summary

Clinical effectiveness

There were relatively few studies in the clinical effectiveness review (12 RCTs) and studies were heterogeneous. They were of small size including a total of around 450 HCL recipients followed for between 4 and 26 weeks, accumulating around 110 person years of observation. Inclusion criteria were relatively narrow and most participants had reasonably good glycaemic control at entry, as indicated in most of those studies reporting baseline time in range (3.9 to 10 mmol/litre) at greater than 50% (range 47% to 62%), and baseline HbA1c at between 7% and 8%.

The NHSE adult pilot study included a broader spectrum of patients with worse glycaemic control at baseline (HbA1c around 9.4%).

Compared with CSII plus CGM, the NMA showed that the HCL arm of the RCTs had a statistically significant improvement (reduction) in HbA1c percentage of -0.29 (95% CI: -0.37 to -0.21). There was a statistically significant increase in percentage time in range between 3.9 to 10 mmol/litre, with a mean difference of 8.6 (7.03 to 10.22). The NMA also showed that percentage time above range (over 10.0 mmol/litre) was significantly decreased, with a mean difference of -7.2 (-8.89 to -5.51). Control arms also showed improvement, but this was less than that seen with HCL. The outcome estimates reported for observational studies were quantitatively

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 45 of 54

broadly in line with those from the RCTs. In the NHSE pilot, transfer to HCL resulted in larger improvements than observed in other studies (decrease in HbA1c of 1.5%), which may be due to the poorer baseline status.

The RCT data suggests that the gains in glycaemic control reported for HCL were not accompanied by a greater risk of hypoglycaemia, however the power to detect small event sizes was limited because of the small size of study groups and relatively short treatment duration.

Adverse events were reported in some studies and were mainly low. Patient reported outcomes were assessed using various methods and did not result in clear trends.

Cost effectiveness

The key model inputs that impacted on results were:

• The net effect upon HbA1c

• The duration of the net effect upon HbA1c

• The model time horizon

• Treatment costs

The modelled cost effectiveness of HCL compared with CSII plus CGM is driven by the change in HbA1c and how long that change persists. It is assumed that the HbA1c effect persists for the patient lifetime, and therefore the baseline age determines the duration of the HbA1c effect. In the base case, the national diabetes audit mean age of those on pumps is used, sampling this using the standard deviation.

With the NMA estimated HCL effect on HbA1c of -0.29% compared with CSII plus CGM, the net total cost was estimated to be £28,628 after accounting for fewer complications (reduced eye and renal complications). There was a net undiscounted survival gain of 0.458 years, contributing to a gain of 0.160 QALYs. This resulted in a base case deterministic ICER of £179k per QALY

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 46 of 54

gained and a probabilistic central estimate of £186k per QALY gained. The EAG said the probability of HCL being cost effective at £20k per QALY and £30k per QALY thresholds were 21% and 31%, respectively.

The ICER was reduced to £126k per QALY gained if the NHSE adult pilot baseline patient characteristics were used. When the NHSE adult pilot change in HbA1c of -1.5% was used this resulted in an ICER of £12,398 per QALY gained. The EAG said that the incidences of renal and eye complications may be overestimated in the model. Adjusting these (that is, reducing by their possible overestimation) increased the ICER to £21,583 per QALY gained.

The modelling of longer term effects was uncertain. Time horizons of 8, 12 and 24 years led to increased ICERs of £910k, £664k and £328k per QALY gained, respectively. The duration of the HbA1c effect was also uncertain. Limiting this to 5, 10 and 20 years while retaining a time horizon of 60 years led to increased ICERs of £657k, £425k and £247 per QALY gained, respectively.

There was high uncertainty around NSHE and SHE annual event rates. There was also a lack of evidence that HCL had an effect on these. When NSHEs were included in a scenario analysis, with an annual rate for HCL of 20.8 (27.1 for CSII plus CGM), the ICER was reduced to £169k per QALY gained. Including SHE’s reduced the ICER further to £163k per QALY gained. Using alternative sources for SHE disutility estimates from Currie et al or Nauck et al, further reduced the ICER to £121k and £109k per QALY, respectively.

Other model inputs used in scenario analyses had a limited effect on the ICER results. These included:

• Doubling the quality of life effect of hypoglycaemia events to reflect possible carer effects.

• Reducing the proportion of CSII plus CGM that is intermittently scanned CGM from 90% to 85%.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 47 of 54

• Increasing the annual cost of HCL systems by £500.

• Applying an additional training cost of £1,132 for transferring from CSII plus CGM to HCL

• Revising non-specific mortality to also exclude deaths due to hypertension or to all-cause mortality.

• Applying an annual 0.045% worsening of HbA1c.

The exploratory modelling of a paediatric population very broadly mirrored the adult results, but the EAG had reservations about the reliability the IQVIA CDM for modelling a paediatric population.

5 Issues for consideration

Clinical effectiveness

Differences in baseline characteristics between RCTs and NHSE pilot led to different estimated HbA1c percentage changes

The network meta-analysis of data from 12 RCTs showed that hybrid closed loop systems were associated with a decrease in HbA1c of 0.29%. The NHSE pilot data showed that hybrid closed loop systems were associated with a decrease in HbA1c levels of 1.50%. Participants in the RCTs had reasonably good glycaemic control at entry (HbA1c between 7% and 8%), whereas the NHSE pilot baseline characteristics included a broader patient base with worse baseline HbA1c levels (around 9.4%). Therefore, this population had greater improvements after HCL treatment (decrease in HbA1c of 1.5%). The EAG used the network meta-analysis result in the base case.

Issues around the RCT and NHSE pilot evidence and generalisability

Clinical effectiveness analysis prioritised RCT evidence. However, the RCTs were of small size with numbers of participants ranging from less than 20 to 135. RCTs were also heterogeneous in terms of trial design, number and age

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 48 of 54

of participants, and other demographics including run-in times, duration of observation periods, and number and types of previous treatments. Three of the RCTs used in the NMA used the Minimed 670G which is an older HCL system and may be expected to result in a smaller reduction/improvement in HbA1c.

NHSE pilot studies were non-randomised studies with no control group and with a before-after study design. The EAG said this could limit the scientific value of the evidence due to greater risk of bias due to lack of randomisation, lack of a true control, and selection bias.

Population subgroup data

In the RCT children and young adults subgroup (under 18 years), the change in HbA1c percentage for HCL was greater (-0.31 [-0.43, -0.20]) than the adult subgroup (-0.24 [-0.32, -0.15]). In the NHSE children and young people pilot, the net HbA1c change was –0.7%. Data was not presented on specific child age groups as were included in the scope (that is, 5 years and under, 6 to 11 years and 12 to 19 years). There was very limited evidence on pregnancy and the effectiveness of HCL in pregnant women remains unclear.

Cost effectiveness

Using the NHSE adult pilot data for HbA1c change results in a large decrease in the ICER

The base case analysis using the NMA estimated HbA1c change resulted in poor cost effectiveness estimates with an ICER of £179k per QALY gained. Using the NHSE adult pilot HbA1c change of -1.5% (along with the pilot baseline patient characteristics) resulted in an ICER of £12,398 per QALY gained.

Differences between rtCGM and isCGM

Most of the clinical evidence had a comparator that used rtCGM, but the model base case assumed 90% isCGM and only 10% rtCGM. Therefore, for

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 49 of 54

the comparator the model is using the clinical effectiveness of rtCGM with the lower cost of isCGM and so may be underestimating the cost-effectiveness of HCL.

The time horizon is a key driver of model results

In the base case the time horizon was 50 years, however modelling of longer term effects is more uncertain. Shorter time horizons explored in scenario analyses resulted in larger ICERs.

Duration of HbA1c effect

The duration of the HbA1c effect is another key driver of the model results. The base case assumes that the effect lasts for the lifetime of the model, however this is uncertain and reducing the duration in scenario analyses also reduces the cost effectiveness. The EAG noted that there is a lack of evidence on the long term effect of the hybrid closed loop system and especially on clinical outcomes such as cardiovascular disease.

Disutilities in the model

There was a lack of data on the effect of HCL on NSHEs and SHEs and also high uncertainty around annual event rates. Therefore, they were not included in the base case. Rates were inferred from the ratio of time below 3.0 mmol/litre for HCL compared to that of the other comparators, coupled with event rates for HCL. The reduction in mental burden and parental or carer anxiety provided by HCL systems may not be captured in the model.

Subgroup modelling

There is uncertainty in the exploratory paediatric modelling results due to the uncertainty around the modelled long term survival coupled with uncertainty about how much of the clinical data used in the IQVIA CDM construction was from a paediatric population. The EAG did not consider the cost effectiveness of HCL for pregnant women due to the lack of evidence.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 50 of 54

6 Equality considerations

NICE is committed to promoting equality of opportunity, eliminating unlawful discrimination and fostering good relations between people with particular protected characteristics and others.

• Some of the hybrid closed loop systems currently available in the UK are not licensed for use in children under 6 or 7 years old and in pregnancy.

• People with certain skin conditions or allergies may be unable to wear a sensor.

• People with learning difficulties and people whose vision or hearing does not allow recognition of pump signals and alarms may have difficulty in using the technologies.

• People who have had diabetes for many years and older people may have impaired awareness of hypoglycaemia.

• There may be a need for tighter glucose control in pregnant women.

• Younger children may need help to operate the device every time and toddlers may have more limited management options.

• People from ethnic minority are less likely to be offered technology as therapy; this may be because of a language barrier.

• People from deprived backgrounds and those who are less educated may be less likely to use the technology; this may be because of less awareness of their options.

• People with cystic fibrosis might be more likely to get diabetes.

• People with blood clotting disorders such as haemophilia might not be able to do finger prick testing.

7 Implementation

CCG funding variation and access

A variation in funding arrangements across Clinical Commissioning groups for continuous glucose monitoring technologies may lead to unequal access to

NICE

Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 51 of 54

technologies for type 1 diabetes. It has been reported that many Clinical Commissioning groups do not have a policy for funding continuous glucose monitoring technologies or have decided not to fund (Choudhary 2019). In addition, some systems are not licensed in certain groups which may limit their options.

Technology requirements

The control algorithms apps for hybrid closed loop systems are typically hosted on smart phones. Some people may use old phones that cannot host these apps or may be unable to buy smart phones, thereby limiting their access to the technology.

System choice and manufacturer support

The choice of system a person prefers may be influenced by the level of support provided by the manufacturer to help resolve technical issues.

DIY closed loop systems

Even though DIY closed loop systems do not have regulatory approval, a growing number of people with type 1 diabetes continue to use these systems. A position statement offering clinical guidance for people who use DIY closed loop technologies has been developed.

8 Authors

Simon Webster

Topic lead

Frances Nixon

Technical adviser

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 52 of 54

Glossary

Bolus

A dose of insulin taken at mealtimes to keep blood glucose levels under control following a meal.

Continuous subcutaneous insulin infusion (CSII)

Continuous subcutaneous insulin infusion is delivered through a subcutaneously inserted cannula connected to an external pump with a refillable storage reservoir. The insulin pump can be tethered, where insulin is sent from the pump to the cannula through a tubing, or patch where the pump is attached directly to the skin.

Diabetic ketoacidosis

An acute short-term complication faced by people with type 1 diabetes when there is insufficient insulin in the body to allow the entry of glucose into cells. This leads to the metabolism of alternative energy sources such as fat, resulting in the harmful build-up of ketones in the blood.

HbA1c

Glycated haemoglobin. It is a measure of average blood glucose levels over the previous 2 to 3 months.

Hybrid closed loop (HCL)

Systems that use a mathematical algorithm to automatically drive insulin delivery in response to real time continuously monitored interstitial fluid glucose levels. They aim to reduce the user and carer input required for insulin monitoring and dosing.

Hyperglycaemia

High blood sugar.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 53 of 54

Hypoglycaemia

Low blood sugar.

Intermittently scanned continuous glucose monitor (isCGM)

Consists of a sensor and transmitter that automatically monitors interstitial fluid glucose levels throughout the day and night but gives glucose readings only when the sensor has been scanned.

Real time continuous glucose monitor (rtCGM)

Consists of a sensor and transmitter that automatically monitors interstitial fluid glucose levels throughout the day and night and sends real time readings to a receiver or smart device.

Time above range

Percentage of time that blood glucose is above 10 mmol/litre and denotes hyperglycaemia.

Time below range

Percentage of time that blood glucose is below 3.9 mmol/litre and denotes hypoglycaemia.

Time in range

A measure of glycaemic control. Usually refers to percentage of time that blood glucose is between 3.9 mmol/litre and 10 mmol/litre.

NICE Evidence overview of Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes [November 2022] Page 54 of 54

Title : Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes

Produced by:

Warwick Evidence

Division of Health Sciences Warwick Medical School, University of Warwick Superseded – see xxxxxxxx, xxx xxx Authors: updated external Asra Asgharzadeh, Mubarak Patel, Martin Connok, Sara Dmery, Iman Ghosh, Mary Jordan, Kevin Momanyi, Karoline Freeman, Rachel Court, Anna Brown, Sharin Baldwin, Fatai Ogunlayi, Chris Stinton, Ewen Cummins, Lena Al-Khudairy Correspondence to: Dr Lena Al-Khudairy assessment report Warwick Evidence, Warwick Medical School, University of Warwick, xxxxxxxx, xxx xxx ve mber Tel: xxx xxxx xxxxx (15 No 2022)

Email: xxxxxxxxxxxxxxxxxxxxxxxxxx

Date completed: 27.09.2022

Source of funding : This report was commissioned by the NIHR Evidence Synthesis Programme as project number NIHR as project number 133547.

Declared competing interests of the authors

None

Acknowledgements

Professor Norman Waugh, Warwick University, who provided clinical and topic support. Professor Aileen Clark for assessing the quality of the work. Dr Nitin Gholap, Consultant in Endocrinology and acute Medicine - University Hospital Coventry and Warwickshire, who provided clinical support. Leigh Carr, NHS supply chain, who provided the costings of pumps and their consumables. Ben McGough, Digital Lead, NHS Diabetes Programme Team, who liaised with the Diabetes Technical Network. Dr. Fiona Green, Consultant in Diabetes and Endocrinology, who provided additional expert opinion. Dr. Mark Lamotte and the iQVIA team who provided advice on inputs to and the use of the iQVIA Core Diabetes Model.

Rider on responsibility for report

The views expressed in this report are those of the authors and not necessarily those of the NIHR Evidence Synthesis Programme. Any errors are the responsibility of the authors. This report should be referenced as follows: Asgharzadeh A., Patel M., Connok M., Dmery S., Ghosh I., Court R., Jordan M, Superseded – see Momanyi K, Freeman K., Brown A., Baldwin S., Ogunlayi F., Stinton C., Cummins E., Al-Khudairy L. Hybrid closed loop systems for managing blood glucose levels in type 1 diabetes. 2022. updated external Contributions of authors Asra Asgharzadeh: senior clinical effectiveness (screening for clinical evidence, assessment of study inclusion, data abstraction, write-up of clinical results) , Mubarak assessment report Patel: statistical analysis (network meta-analysis, write-up of results), Martin Connok: statistical analysis (clinical study analysis, write-up of results), Sara Dmery: clinical effectiveness (screening for clinical evidence, assessment of study inclusion, data abstraction, risk of bias write-up), Iman Ghosh: clinical effectiveness (quality appraisal), Mary Jordan: systematic review of cost-effectiveness, Kevin Momanyi: (15 November 2022) screening for clinical evidence, assessment of study inclusion, quality assessment, Karoline Freeman: clinical effectiveness (screening for clinical evidence, assessment of study inclusion, data abstraction, Rachel Court: senior information specialist (design and update searches, locate records, reference), Anna Brown: information specialist (design and update searches, locate records), Sharin Baldwin: clinical effectiveness (screening for clinical evidence, assessment of study inclusion), Fatai Ogunlayi: clinical effectiveness (screening for clinical evidence, assessment of study inclusion), Chris Stinton: clinical effectiveness (screening for clinical evidence, assessment of study inclusion), Ewen Cummins: senior cost-effectiveness (costeffectiveness lead and production), Lena Al-Khudairy: clinical effectiveness lead.

Please note that: Sections highlighted in yellow and underlined are ‘academic in confidence’ (AIC). Sections highlighted in aqua and underlined are ‘commercial in confidence’ (CIC). Figures that are CIC have been bordered with blue. Depersonalised Data (DPD) is highlighted in pink.

ABSTRACT

Background : Hybrid closed loop systems are a new class of technology to manage type 1 diabetes. The system includes a combination of real-time glucose monitoring from a continuous glucose monitoring device and a control algorithm to direct insulin delivery through an insulin pump. Evidence suggest that such technologies have the potential to Superseded – see improve the lives of people with type 1 diabetes and their families. Aim: The aim of this appraisal was to assess the clinical and cost effectiveness of hybrid closed loop systems for managing glucose in people who have T1DM, and are having updated external difficulty managing their condition despite prior use of at least one of the following

technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring. assessment report Methods: a systematic review of clinical and cost-effective evidence following a predefined inclusion criteria informed by the aim of this review. An independent economic assessment using iQVIA CDM to model cost effectiveness. Results: The clinical evidence identified 12 randomised controlled trials (RCTs) that (15 November 2022)

compared HCL to CSII+CGM or SAP therapy. HCL arm of RCTs achieved improvement in HbA1c % (HCL decreased HbA1c % by 0.28 (-0.34 to -0.21), increased % TIR (between 3.9 – 10.0 mmol/L) with a mean difference of 8.6 (7.03 to 10.22), significantly decreased TIR (% above 10.0 mmol/L), with a mean difference of -7.2 (-8.89 to -5.51) but did not significantly affect % time within range (<3.9 mmol/L). Comparator arms also showed improvements but this was less than that observed in the HCL arm. Outcomes were superior in the HCL arm vs. comparator arm. The cost effectiveness search identified six studies which were included in the review systematic

review. Studies reported subjective cost-effectiveness that was influenced by the willingness to pay thresholds. Economic evaluation showed that the published model validation papers suggest that an earlier version of the iQVIA CDM tended to overestimate the incidences of the complications of diabetes, this being particularly important for severe visual loss and ESRD. Medium term modelling of overall survival appeared good, but there was uncertainty about its longer term modelling.

3

Current prices suggest that HCL is around an annual average £1,500 more expensive than

CSII+CGM, though this may increase by around a further £500 for some systems.

The EAG base case applies the EAG RCT NMA estimate of -0.29% HbA1c for HCL relative to CSII+CGM. There was no direct evidence of an effect upon symptomatic or severe hypoglycaemia events, therefore the EAG does not include these in its base case. Superseded – see The change in HbA1c results in a gain in undiscounted life expectancy of 0.458 years and a gain of 0.160 QALYs. Net lifetime treatment costs are £31,185, with reduced complications leading to a net total cost of £28,628. The cost effectiveness estimate is updated external £179k per QALY. The EAG has some concerns about using the iQVIA T1DM to model a paediatric population. The EAG does not formally consider the cost effectiveness of HCL compared to CSII+CGM for pregnant women. It only notes the relationship between HbA1c and birth defects. assessment report

Conclusions: RCTs of HCL interventions in comparison CSII+CGM or sensor augmented pump therapy achieved a statistically significant improvement in HbA1c %, in TIR between 3.9 to 10 mmol/L, and in hyperglycaemic levels. (15 November 2022) Word count: 526

4

SCIENTIFIC SUMMARY

Background

Type 1 diabetes was formerly known as insulin-dependent diabetes. It is the result of an autoimmune process leading to destruction of the insulin-producing beta cells in the pancreas. The cause of this auto-immune disease is not known. Diabetes is managed by Superseded – see lifestyle and education, glucose monitoring, and insulin delivery. Treatment with insulin is aimed at replicating the function of the pancreas. The aim of treatment is to control hyperglycaemia and avoid hypoglycaemia. The NICE target for type 1 diabetes is 48 updated external mmol/mol (formerly 6.5%) but few people with T1DM achieve that. Interventions to manage diabetes include: education, continuous glucose monitoring (include a sensor, transmitter and display device), insulin therapy (multiple daily injections or continuous assessment report subcutaneous insulin infusion). Continuous subcutaneous insulin infusion (CSII) is an alternative therapy to multiple daily injections. CSII is an external pump that delivers insulin continuously from a refillable storage reservoir by means of a subcutaneously placed cannula. Sensor-augmented pump (SAP) therapy systems combine CGM with (15 November 2022)

CSII. The systems are designed to measure interstitial glucose levels (every few minutes) and allow immediate real‑time adjustment of insulin therapy. The systems may produce alerts if the glucose levels become too high or too low. SAP can operate in standard (manual) and advanced (automatic) modes. In the manual open loop mode, the continuous glucose monitor and glucose pump do not communicate with each other, and insulin doses are programmed by the user, who makes manual adjustments. Hybrid closed loop systems are a new class of technology that use a combination of real-time glucose monitoring from a continuous glucose monitoring device and a control algorithm to direct insulin delivery through an insulin pump. Evidence suggest that such technologies have the potential to improve the lives of people with type 1 diabetes and their families.

Objectives

The intervention of interest is a class of automated insulin delivery systems which consists of three components – a CGM, a microprocessor with control algorithms, and a pump. The

5

overall objectives of this project are to examine the clinical and cost-effectiveness of hybrid closed loop systems for managing glucose levels in people who have T1DM.

1. What is the clinical effectiveness of hybrid closed loop systems for managing

glucose in people who have T1DM and are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous Superseded – see subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring? 2. What is the cost effectiveness of hybrid closed loop systems for managing glucose updated external in people who have T1DM, and are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring? assessment report Methods Systematic review methods followed the principles outlined in the Cochrane Handbook of Diagnostic Test Accuracy and the NICE Diagnostic Assessment Programme manual. (15 November 2022) A comprehensive search was developed iteratively and undertaken in a range of relevant bibliographic databases and other sources, following the recommendations in Chapter 4 of the Cochrane Handbook for Systematic Reviews of Interventions. Date limits have been used, in order to identify records added to databases since the searches for DG21 (run in 2014). Two reviewers screened titles and abstracts and assessed eligibility of studies. Studies that satisfy the following criteria were included:

Populations: People who have T1DM who are having difficulty managing their condition

despite prior use of at least one of the following technologies: continuous subcutaneous insulin

infusion, real time continuous glucose monitoring, flash glucose monitoring

If evidence permits the following T1DM subpopulations will be included:

• Pregnant women and those planning pregnancies (excluding gestational diabetes).[b]

• Children (5 years and under, 6 – 11 years, 12 - 19 years).

6

• People with extreme fear of hypoglycaemia.

People with diabetes related complications that are at risk of deterioration.

Target: Type 1 diabetes mellitus

Intervention: Hybrid closed loop systems Superseded – see

Comparator: Real time continuous glucose monitoring with continuous subcutaneous insulin

infusion (non-integrated). Intermittently scanned (flash) glucose monitoring with continuous subcutaneous insulin infusion. updated external Outcomes: Intermediate measures

• Time in target range (percentage of time a person spends with blood glucose level in target range of 3.9-10 mmol/l)

• assessment report • Time below and above target range

• Change in HbA1c

• • Rate of glycaemic variability • Fear of hypoglycaemia

• (15 November 2022) • Rate of severe hypoglycaemic events

  - Rate of severe hyperglycaemic events 
  
  - Episodes of diabetic ketoacidosis 
  
  - Rate of ambulance call outs 
  
  - Rate of hospital out-patient visits 
  
  - Rate of weight gain
  

Clinical outcomes

• Retinopathy

• Neuropathy

Intermediate measures

• Time in target range (percentage of time a person spends with blood glucose level in target range of 3.9-10 mmol/l)

• Time below and above target range

7

• Change in HbA1c

• Rate of glycaemic variability

• Fear of hypoglycaemia

• Rate of severe hypoglycaemic events Superseded – see • Rate of severe hyperglycaemic events • Episodes of diabetic ketoacidosis • Rate of ambulance call outs updated external • Rate of hospital out-patient visits • Rate of weight gain asse Clinical outcomes ssment report • Retinopathy • Neuropathy (15 November 2022) • Cognitive impairment

• End-stage renal disease

• Cardiovascular disease

• Mortality

Additional clinical outcomes in women who are pregnant/have recently given birth:

• Premature birth

• Miscarriage related to fetal abnormality

• Increased proportion of babies delivered by caesarean section

• Macrosomia (excessive birth weight)

• Respiratory distress syndrome in the new-born

Device related outcomes

• Adverse events related to the use of devices

8

Patient-reported outcomes

• Heath-related quality of life

• Psychological well being • Impact on patient (time spent managing the condition, time spent off work or school, ability to participate in daily life, time spent at clinics, impact on sleep) Superseded – see • Anxiety about experiencing hypoglycaemia • Acceptability of testing and method of insulin administration ed external Carer reported outcomes updat • Impact on carer (fear of hypoglycaemia, time spent managing the condition, time spent off work, ability to participate in daily life, time spent at clinics, impact on sleep) assessment report Study design: Hybrid closed loop systems studies included any design. All comparator studies: comparative effectiveness studies. Healthcare setting: Self-use supervised by primary or secondary care (15 November 2022)

Publication type: Peer reviewed papers

Language: English

Prioritization for full text assessment: We applied a two-step approach for identifying and assessing relevant evidence. The elements used to prioritise evidence (study design, study length, sample size). The most rigorous and relevant studies (mainly RCTs) were prioritised for data extraction and quality assessment. Observational studies were

recorded and reported narratively. Two reviewers extracted data independently, using a piloted data extraction form. Disagreements was resolved through consensus, with the inclusion of a third reviewer when required. The risk of bias of randomised trials was assessed using the revised Cochrane risk-of-bias tool for randomized trials. We synthesised the evidence statistically. The network meta-analysis was conducted under a frequentist approach using a random-effects model.

Results

9

Clinical

Systematic review

The clinical evidence identified 12 randomised controlled trials that compared HCL to CSII+CGM or SAP therapy. Studies were heterogeneous in terms of population, age groups, gender, RCT design (parallel cross over), numbers of participants and variable Superseded – see adjustment methods for determining mean difference between intervention and comparators. Studies did not consistently describe comparators. Cross-over studies did not provide data at different cross-over time points. Overall, the HCL arm of RCTs updated external achieved improvement in HbA1c % (HCL decreased HbA1c % by 0.28 (-0.34 to -0.21), increased % TIR (between 3.9 – 10.0 mmol/L) with a mean difference of 8.6 (7.03 to 10.22), significantly decreased TIR (% above 10.0 mmol/L), with a mean difference of - 7.2 (-8.89 to -5.51) but did not significantly affect % time within range (<3.9 mmol/L). assessment report Comparator arms also showed improvements but this was less than that observed in the HCL arm. Outcomes were superior in the HCL arm vs. comparator arm. Available evidence from the RCTs suggests that these gains in glycaemic control reported for HCL (15 November 2022) were not accompanied by a greater risk of hypoglycaemia however the power to detect

small event sizes was limited because of small size of study groups and relatively short treatment duration.

External submissions

NHSE submitted two observational audit studies, the first audit was conducted in adults and the second in children and young people (CYP). The audit included adult

participants that had XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

10

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXX. e Superseded – se Economics Systematic literature review of cost effectiveness The literature search identified six studies which were included in the review systematic updated external review. Five of these studies were economic evaluations of hybrid closed loop systems, whereas one was a budget impact analysis that aimed at estimating the financial impact of reimbursing HCL systems for individuals with type 1 diabetes. These studies were assessment report assessed using the CHEERS and Phillips checklists where applicable. According to the assessment, four studies were identified as cost effectiveness analyses in their titles The structure of the models used in the cost effectiveness studies was judged to be of good quality. The studies clearly stated their decision problem/research question, the viewpoint (15 November 2022)

of their analyses and their modelling objectives, which were coherent with the decision problem. Both the IQVIA CORE Diabetes Model and the Sheffield type 1 diabetes model are validated models for evaluating diabetes technologies. The studies that used the IQVIA CORE diabetes Model described the model as one with a complex semiMarkov model structure with interdependent sub-models, so more thorough, easier access to its reported features would be of benefit to the intended audience. None of the studies clearly showed the illustrative model structure, which depicted the clinical pathway for T1DM. All the cost effectiveness studies noted that hybrid closed loop systems were cost effective over the lifetime compared with their comparator interventions. This inference was, however, subjective as the studies chose arbitrary willingness to pay thresholds. A major limitation of most of the cost effectiveness studies is that their findings might not be generalisable. This is because the studies did not use baseline characteristics and treatment effects data for their target populations.

Company submission

11

The EAG received economic submissions from Medtronic, Dexcom and Camdiab. The Tandem submission referenced the economics of the Dexcom submission.

The Medtronic treatment costs applied the anticipated April 2023 CiC prices rather than

current list prices. Using the iQVIA CDM it estimated that compared to the 640G system with rtCGM the 780G HCL system improved HbA1c by 0.8% which resulted in a saving Superseded – see of £5,816, patient gains of 0.21 QALYs and dominance for HCL. For the comparison with CSII+isCGM the same HbA1c improvement was applied alongside an annual reduction of 0.9 severe hypoglycaemia events. This resulted in a net cost of £13,057, a patient gain of 0.70 QALYs and a cost effectiveness of £18,672 per QALY. updated external

Dexcom used the XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX a XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX rt ssessment repo XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXX XX XX XXXXX XXX (1 5 Nove XXXXXXXX XX XX XXXX mber 202 2) The Camdiab submission presented XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX. XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXX

12

Independent economic assessment

Due to the complexity of modelling T1DM the EAG does not build a de novo model.

There are two main T1DM economic models available, the Sheffield T1DM model and the iQVIA CDM. In common with NG17 and DG21 and most of the company submissions, the EAG uses the iQVIA CDM to model cost effectiveness. The published Superseded – see model validation papers suggest that an earlier version of the iQVIA CDM tended to overestimate the incidences of the complications of diabetes, this being particularly important for severe visual loss and ESRD. Medium term modelling of overall survival appeared good, but there was uncertainty about its longer term modelling. It is not known updated external whether these issues persist in the current iQVIA CDM. The EAG assesses the cost effectiveness of HCL, PLGS and CSII+CGM. PLGS is extendedly dominated throughout and for this summary the EAG does not consider it assessment report further. Direct treatment costs are supplied by the NHS supply chain using current list prices. The EAG provides a cPAS appendix that applies the confidential possible future prices. (15 November 2022)

Current prices suggest that HCL is around an annual average £1,500 more expensive than

CSII+CGM, though this may increase by around a further £500 for some systems.

CSII+CGM is cheaper than HCL in large part due to 90% or more of adult patients using isCGM sensors rather than rtCMG sensors.

Patient baseline characteristics for the EAG base case are drawn from the National Diabetes Audit subgroup of T1DM patients on pumps.

The EAG base case applies the EAG RCT NMA estimate of -0.29% HbA1c for HCL relative to CSII+CGM. Due to there being no direct evidence of an effect upon symptomatic or severe hypoglycaemia events the EAG does not include these in its base case.

The change in HbA1c results in a gain in undiscounted life expectancy of 0.458 years and a gain of 0.160 QALYs. Net lifetime treatment costs are £31,185, with reduced complications leading to a net total cost of £28,628. The cost effectiveness estimate is £179k per QALY.

13

The EAG provides scenario analyses that estimate symptomatic and severe

hypoglycaemia events based upon the differences in the time below 3.0mmol/l for HCL

and CSII+CGM. These improve the cost effectiveness of HCL to £163k per QALY if

valued using the EAG preferred source, to £121k if valued using the same source as NG17 and to £109k if valued using other credible sources. Superseded – see These results show are sensitive to time horizons of less than the patient lifetime, durations of HbA1c effect of less than the patient lifetime and higher HCL treatment costs which tend to worsen the cost effectiveness of HCL. If mortality for those without complications is higher than that of the base case or there is an annual worsening of updated external HbA1c this tends to improve the cost effectiveness of HCL. All the resulting cost assess effectiveness estimates are above £100k per QALY. If the NHSE adult pilot change XXXXXXXXXXXXXXXXX XXXX XXXXX ort is ment rep assumed to be the net effect of HCL compared to CSII+CGM the undiscounted gain in life expectancy more than doubles to 1.004 years, and the patient gain to 3.103 QALYs. Net lifetime treatment costs increase to £35,912 due to the greater life expectancy, but (15 November 2022)

considerable cost savings from reduced eye complications of £16,442 and reduced renal

complications of £6,731 lead to a net total cost of £12,447 and a cost effectiveness of

£12,398 per QALY. Reducing the modelled complication costs by their possible

overestimation worsens the cost effectiveness to £21,583 per QALY. This does not take

into account any quality of life effects and survival effects from possible overestimation of complication rates.

The key model inputs are:

• The net effect upon HbA1c.

• The duration of the net effect upon HbA1c.

• The model time horizon.

• Treatment costs.

Other important model inputs are:

• Hypoglycaemia event rates.

• What source is used to value the disutilities of hypoglycaemia event rates.

14

• What non-specific mortality is applied.

• Whether HbA1c worsens annually among T1DM patients and if so by how much.

The key modelling uncertainties are around: • Overall survival gains. • Severe visual loss and its effects upon survival, quality of life and costs. Superseded – see • ESRD and its effects upon survival, quality of life and costs. updated external The EAG has some concerns about using the iQVIA T1DM to model a paediatric population. Exploratory modelling of a paediatric population broadly mirrors that of the adult population, though the NHSE paediatric pilot reported XXXXXXXXXXXXX assessment change between baseline and six months with a corresponding XXXXXXXX in the cost o rt rep effectiveness estimate for this scenario. The EAG does not formally consider the cost effectiveness of HCL compared to (15 November 2022)

CSII+CGM for pregnant women. It only notes the relationship between HbA1c and birth

defects. If HCL reduces HbA1c in pregnant women to the same extent as in the adult population the short-term additional costs of HCL will have some immediate cost offsets from reduced birth defects, with the potential for additional benefits to the child at no additional cost. It also seems likely that the baseline age of pregnant women is below the national diabetes audit mean age which is likely to further improve cost effectiveness. If after giving birth women remain on HCL into the long term the cost effectiveness estimate of HCL may trend towards that of the adult female T1DM population of the same age, but will remain superior to it.

Conclusions

RCTs of HCL interventions in comparison CSII+CGM or sensor augmented pump therapy achieved a statistically significant improvement in HbA1c %, in TIR between 3.9 to 10

15

mmol/L, and in hyperglycaemic levels. The outcome estimates reported for observational studies were quantitatively broadly in line with those from the RCTs. Measures of glycaemic performance such as HbA1c%, % time in range (3.9 to 10 mmol/L), and % time above range >10 mmol/L all improved on transfer to HCL. There is a research need of well designed studies because identified studies were heterogeneous in terms of population, age Superseded – see groups, gender, RCT design (parallel cross over), numbers of participants and variable adjustment methods for determining mean difference between intervention and comparators. Future research should clearly describe comparators because this is not clear in the current literature. updated external Word count: 3182 assessment report (15 November 2022)

PLAIN ENGLISH SUMMARY

16

Type 1 Diabetes (T1DM) is a life-long condition where the individual’s pancreas significantly reduces \ stops producing the hormone insulin that manages blood glucose levels. As a result, the individual must self-administer insulin, monitor their blood glucose

levels, and take into consideration many multiple variables to achieve a tight blood glucose control range. Superseded – see With the challenge of self-management, blood glucose levels may swing high (hyperglycemia) and low (hypoglycemia) multiple times a day. This can result in the individual experiencing confusion, fatigue, nausea and possible unconsciousness as part of updated external their daily management. The long-term risks of high blood glucose levels include damage

to blood vessels, impacting sight, sense of touch and other vital organs. During self management, the individual uses the information they have to administer the amount of assessment report insulin the body requires while limiting high and low blood sugar. The day-to-day

management of diabetes can be difficult and, and at times people with diabetes may struggle to maintain control of their blood glucose level. This can put a significant burden on the patient and carers which can result in impact on quality of life and a feeling that the (15 November 2022)

condition limits \ controls their abilities.

Management of Type 1 Diabetes

Type 1 Diabetes is managed via lifestyle adjustments and review of multiple sources of data to help calculate the amount of insulin that a person needs. This commonly covers the following:

● Lifestyle

• A balanced diet including complex carbohydrates, fats and proteins and avoiding processed food slows the impact of food on the blood glucose level reducing the possibility of sudden highs or lows.

• Exercise improves the body's sensitivity to insulin, therefore, reducing the amount to be injected. This can reduce the possibility of unexpected sudden blood glucose changes that a larger dose of insulin may bring, as well as general well-being in reducing stress that can cause insulin resistance.

17

• Data

  • Patients' understanding and monitoring of their body’s reaction to insulin

and foods to calculate their sensitivity to insulin and carbohydrates.

○ Monitoring of blood glucose levels via “finger pricks” where the individual draws a small amount of blood to get a point in time reading or continuous Superseded – see glucose monitors that provide a real-time reading of blood glucose. ● Insulin Delivery ○ Via daily injections or insulin pump that is connected to the body 24/7. Injections can be of rapid acting insulins that take effect within a short time updated external frame (bolus) and long-acting insulins that release over a 12-to-24 hour period providing an amount of background insulin in the body (basal). Insulin pumps provide rapid acting insulin with the ability to deliver a bolus assessment report quickly and easily along with continuous background basal delivery that can be precisely adjusted for example every 5 minutes to form a unique 24-hour profile for the individual. (15 November 2022)

Processing of this information and deciding the best action is an ongoing challenge for the individual. Examples of such challenges include:

• Diet: Poor diet education, cost of access to fresh food and the challenge of avoiding easily accessible but cheap highly processed foods.

• Exercise: Lifestyle habits and motivation to exercise, along with the management of changes to insulin sensitivity, during and after exercise.

• Insulin Delivery: The inconvenience of injections and their limited control of insulin delivery, pumps with an overwhelming number of options for consideration.

• Blood Glucose Monitoring: This can be uncomfortable and provide a person with limited visibility of trend data. Compared to the data provided by manual blood glucose tests, continuous glucose monitors provide an overwhelming amount of real-time data for the individual to process.

18

• Alarm fatigue: insulin pumps can cause frustration, due to automatic alarms set to inform the individual of high or low blood glucose or lack of proactive information to prevent such events.

● Overtreatment: Miscalculation, frustration or unexpected sensitivity/resistance to insulin that can result in multiple blood sugar highs and lows within a short Superseded – see timeframe. ● Changes in sensitivity to insulin, and to food along with many other factors that can changes an individuals response to insulin over time and day to day. updated external Hybrid closed loop systems Hybrid closed loop systems provide a control algorithm that reviews data, along with reviewing the impact of its past actions. It can action frequent minor adjustments of insulin assessment report delivery to allow blood glucose levels to be managed. The system is proactive versus reactive using the real-time feed of data provided by the continuous glucose monitor to make calculations and take actions and to take actions using a high level of controlled (15 November 2022)

delivery offered by an insulin pump at a frequency that is unattainable by a human being.

As a result, such systems can significantly reduce the burden on the patient by taking responsibility for handling the volume of data and technology required for management of their condition and providing intervention when needed.

The aim of the current project is to review the clinical and cost-effectiveness of hybrid closed loop systems for managing glucose in people who have T1DM and are having difficulty managing their condition.

19

TABLE OF CONTENTS

20

21

22

23

24

LIST OF TABLES AND LIST OF FIGURES

==> picture [569 x 593] intentionally omitted <==

----- Start of picture text -----

----- End of picture text -----

25

==> picture [569 x 324] intentionally omitted <==

----- Start of picture text -----

----- End of picture text -----

26

1 DEFINITION OF TERMS AND LIST OF ABBREVIATIONS

27

28

29

AHCL Advanced Hybrid Closed Loop UADE Unanticipated adverse device effects WTP Willingness to pay Superseded – see updated external assessment report (15 November 2022)

30

2 BACKGROUND

2.1 Description of health problem

Type 1 diabetes was formerly known as insulin-dependent diabetes. It is the result of an autoimmune process leading to destruction of the insulin-producing beta cells in the pancreas. Superseded – see The cause of this auto-immune disease is not known. 2.1.1 Aetiology, pathology and prognosis Insulin is essential for survival. Diabetes is characterised by high blood glucose levels – updated external hyperglycaemia. Injected insulin lowers blood glucose. It can cause abnormally low glucose

– hypoglycaemia. The aim of insulin treatment is to keep plasma glucose as close to normal as possible and so prevent the development of the long-term complications of diabetes due to assessment report hyperglycaemia, including • retinopathy, which can lead to visual impairment and blindness • nephropathy which can lead to renal failure and dialysis (15 November 2022)

• neuropathy, which can cause various symptoms and increase the risk of amputation

Treatment also aims to reduce the increased risk of cardiovascular disease seen in diabetes. Deficiency of insulin can lead to diabetic ketoacidosis which can be fatal.

2.1.2 Epidemiology

Type 1 diabetes usually comes in late childhood or early adolescence but can develop at any age. Type 1 diabetes accounts for 5-10% of diabetes cases. The prevalence of type 1 diabetes is higher in adults than in children, the highest prevalence is observed in adults aged 30 years and above.[1, 2] There are about 250,000 people with T1DM in the UK.

2.1.3 Impact of health problem

Hypoglycaemia

Hypoglycaemia can be mild, moderate or severe.

People with diabetes are rightly scared of hypoglycaemia, and this fear may lead to them allowing blood glucose to run higher than is desirable which can increase the risk of longterm complications. The episodes of hypoglycaemia are usually called “hypos”.

The American Diabetes Association[3] defines hypoglycaemia as follows;

31

• 1. Severe hypoglycemia: an event requiring assistance of another person to actively

  • administer carbohydrate, glucagon, or other resuscitative actions. These episodes may

  • be associated with sufficient neuroglycopenia to induce seizure or coma.

2. Documented symptomatic hypoglycemia: an event during which typical symptoms of hypoglycemia are accompanied by a measured plasma glucose concentration of 3.9 Superseded – see mmol/l). 3) Asymptomatic hypoglycemia: an event not accompanied by typical symptoms of hypoglycemia but with a measured plasma glucose concentration of 70 mg/dl (3.9 updated external mmol/l). Non-severe hypoglycaemia can be mild or moderate. Mild hypoglycaemia may present with symptoms such as sweating, shaking, hunger, and nervousness. Some symptoms are due to the release of adrenaline. Mild is easily self-managed by taking rapidly-absorbed assessment report carbohydrate. Moderate hypoglycaemia can cause difficulty concentrating or speaking, confusion, weakness, vision changes and mood swings. (15 November 2022)

Mild and moderate hypos can usually be managed by the diabetic person themselves, but moderate hypos often lead to interruption of activities.

In the guidance on the Medtronic Veo suspend pump (DG21), NICE defined disabling hypoglycaemia as follows:

“People with type 1 diabetes may experience 'disabling hypoglycaemia', which is when hypoglycaemic episodes occur frequently or without warning so that the person is constantly anxious about having more episodes. This can have a negative effect on quality of life.”

Severe hypoglycaemia can lead to cognitive impairment, unconsciousness and convulsions, and can be fatal. People having severe hypos need assistance and may need to attend an accident and emergency (A&E) department, seek support from paramedics. They may require admission to hospital. A population-based study in (2003) by Leese and colleagues[4] in Tayside found that on average, about 1 person in 14 had a hypo event each year which was severe enough to require NHS assistance, from the ambulance service, A&E, or admission. In young children, repeated severe hypos can cause some cognitive impairment.

Hypoglycaemia can trigger an adrenergic response that acts as a warning that glucose should be consumed. Unfortunately, in some people, after repeated hypos, this warning may be lost.

32

This is known as hypoglycaemic unawareness, and such people are at increased risk of severe

hypoglycaemia and its effects. These individuals are covered by the recommendation in DG21[5] and in TA151,[6] in guidance on insulin pumps.

Nocturnal hypoglycaemia occurs during sleep and may not be detected. However it may disturb sleep and wake people up. It can have two adverse effects. One is rebound Superseded – see hyperglycaemia, the result of the body’s reaction to hypoglycaemia such as release of other

hormones that increase blood glucose, so that nocturnal hypoglycaemia may result in unusually high blood glucose levels around breakfast. The other consequence is that nocturnal hypoglycaemia may itself contribute to hypoglycaemic unawareness. updated external Past appraisals In a technology appraisal (TA53) of long-acting insulin analogues (at that time only glargine),[7] the NICE Appraisal Committee accepted that both hypoglycaemic episodes, and assessment report the fear of such episodes recurring, caused significant disutility. A utility decrement of 0.0052 per non-severe hypoglycaemic event (NSHE) was accepted. As regards fear of hypos, the NICE Glargine guidance (TA53)[7] states: (15 November 2022)

“ The Committee accepted that episodes of hypoglycaemia are potentially detrimental to an individual’s quality of life. This is partly the result of an individual’s objective fear of symptomatic hypoglycaemic attacks as indicated in the economic models reviewed in the Assessment Report. In addition, as reported by the experts who attended the appraisal meeting, individuals’ quality of life is affected by increased awareness and uncertainty of their daily blood glucose status and their recognition of the need to achieve a balance between the risk of hypoglycaemia and the benefits of longer-term glycaemic control. The Committee understood that improvement in this area of concern regarding the balance between hypoglycaemia and hyperglycaemia could have a significant effect on an individual’s quality of life.”

However, the guidance did not specify the amount of utility lost because of fear of hypos, and nor did the Technology Assessment Report[8] because it was based on the industry submission from Aventis, which was classed as confidential. But clearly the utility gain from reducing the fear of hypoglycaemia was enough to change a substantial cost per QALY to an affordable one. There is the probability that a reduction in the rate of severe hypoglycaemia events may reduce the fear of severe hypoglycaemia events, though the impact of this seems likely to be variable across patients. The quality-of-life impact arising from this would be

33

over and above the direct quality of life impact of the severe hypoglycaemia events in

themselves.

Fear of severe hypos was estimated to reduce QoL by 0.020 in the development of type 2

guidelines in 2008. The assessment group (Waugh et al, Aberdeen[9] ) considered the reasonableness of this Superseded – see “This fear effect may only apply to a sub-group of patients, but as an illustration of the possible impact of this, the social tariffs derived by Dolan and colleagues[10] suggest that a move from level 2 within the anxiety subscale of EQ-5D to level 1 would be associated with a 0.07 QoL gain. In a similar vein, the coefficients derived by Brazier and colleagues updated external[11] for the SF-6D questionnaire for the consistent model using standard gamble valuations suggest that a movement within the social dimension from health problems interfering moderately to not interfering would be associated with a 0.022 QoL improvement. Similarly, an assessment report improvement in the mental health subscale from feeling downhearted some of the time to little or none of the time would be associated with a 0.021 QoL improvement.” Studies of the disutility of hypoglycaemia (15 November 2022)

Brod et al[12] carried out a survey to estimate the effect of non-severe hypos on work – productivity, costs and a self-management. They used telephone interviews and focus groups, supplemented by a literature review. Respondents were required to have had a non-severe hypoglycaemic event (NSHE) in the previous month. NSHE was defined as a hypo event not requiring assistance from anyone else, with or without blood glucose measurement, and with or without symptoms. They were asked about duration, effect on work, and likely cause, and whether it occurred at work, at other times of day, or during sleep. 713 had type 1 diabetes, and half of this group had NSHEs at least once a week, with 27% having at least one a month. 22% had hypos only a few times a year.

About 95% of people identified hypos by symptoms, and about 60% of episodes were confirmed by a blood glucose test. The average duration of a NSHE was 33 minutes, but the effect on self-management lasted a week, with an extra six blood glucose tests, a reduction in insulin dose by an average of 6.5 units per day for 4 days in 25% of people, and an unplanned contact with a health care professional by 25%.

The effects on work included;

• Leaving early or missing a full day in 18%. The average work time lost was 10 hours.

34

• Missing meetings or being unable to finish a task – 24%

Work time was lost not only because of NSHEs occurring at work but also outwith work

including nocturnal hypos. No breakdown by insulin regimen was reported such as CSII versus MDI. Leckie et al[13] recruited 243 people with diabetes (216 people with T1DM and some with Superseded – see T2DM on insulin) who were in employment. Their insulin regimens included mostly MDI but 51 were on twice-daily mixtures of soluble and NPH. Over a 12-month follow-up, they recorded their hypo events, severity and effect on work, every month. A total of 1,955 NSHEs were reported, plus 238 severe hypos (some involving unconsciousness and seizures, updated external and a few resulted in soft tissue injuries). However, 66% of patients had no severe hypos. Most (62%) of the severe episodes occurred at home, 52% during sleep, but 15% occurred at work. 55% of the NSHEs occurred at home and 30% at work. It should be noted that the assessment report mean HbA1c was over 9% in most patients, with the exception of patients having more than two severe hypos over the year, in whom it was 8.4% - still far above target. Frier et al[14] carried out a survey amongst 466 people with T1DM of the frequency of non(15 November 2022) severe hypoglycaemia and found that people with T1DM had an average of 2.4 episodes a week (median = 2), with around a quarter being nocturnal. The after-effects include fatigue and reduced alertness, and persisted longer after nocturnal NSHEs (10 hours) than after

daytime episodes (5 hours). Amongst those in employment, 20% of NSHE led to loss of work time. Most did not contact their health care professionals. Self-testing of blood glucose increased in the week after the episode, with an average 4 extra tests. The survey showed that NSHEs are troublesome for patients and have effects lasting at least into the following day. The commonest after-effects were tiredness, reduced alertness and feeling emotionally down. Choudhary et al[15] reported that use of pumps with a low glucose suspend facility meant that 66% of NSHEs lasted less than 10 minutes, and only 12% lasted for up to 2 hours. Nocturnal hypos were greatly reduced.

About 30% of people with type 1 diabetes have impaired awareness of hypos[16] and they are 3-6 times more likely to have severe hypos. The Gold scale rates awareness on a scale of 1 to 7 where 7 means complete absence of symptoms of hypoglycaemia. Structured education such as DAFNE restores awareness in about half of people with impaired awareness. Better control with avoidance of hypoglycaemia can also restore awareness. A trial by Little et al[17] (the HypoCOMPass trial) showed that better control for 24 weeks improved the Gold score

35

by one point and reduced the fear of hypo level from 58 to 45 (higher scores indicate greater

fear, with the maximum being 132), without adversely affecting HbA1c.

Evans et al[18] used the time trade-off method to estimate the disutility of hypos on the

HRQoL scale (0 to 1 where I is perfect health and 0 is death). They interviewed 551 people with type 1 diabetes and 8286 people with no diabetes. They note that hypos can affect Superseded – see HRQoL in two ways, firstly the direct effects of the episodes, and secondly through fear of

future hypos which can lead to precautions such as insufficient insulin dose (increasing the risk of complications), restricting physical activity, over-eating. In addition, repeated hypos can lead to hypoglycaemic unawareness which increases the risk of future hypos. They updated external

estimated that daytime NSHEs reduce HRQoL in a range of 0.032 for one event a month to 0.071 for three episodes a week. Nocturnal NSHEs reduce it by slightly more. Severe events, even only once or twice a year, reduce HRQoL by about 0.08. assessment report

The general public valuation of disutility per event per year ranged from 0.004 for non-severe

daytime hypos to 0.06 per severe event. People with type 1 diabetes had slightly lower estimates of the disutility of severe events, at 0.047. (15 November 2022) Using data from this study, Lauridson et al[19] reported that the disutility of NSHEs may diminish if there are repeated events.

The study by Harris et al[20] reports the Canadian results from this study.

Levy and colleagues[21] elicited utility values for non-severe hypoglycaemia from 51 Canadians (but only half had T1DM) and non-diabetic controls. The disutility from a single NSHE was 0.0033. Levy et al argue that a minimum significant utility loss is 0.03, which would be reached by people having 10 NSHEs a year.

Adler et al[22] found that severe, frequent and nocturnal hypoglycaemia reduced quality of life, ranging from 0.84 in people with diabetes who had the least severe state) non-severe, daytime only, only once a year, not causing any worry) to 0.40 (severe frequent hypoglycaemia day and night, causing anxiety).

Currie and colleagues[23] surveyed 1,305 UK patients with type 1 and type 2 diabetes using both the Hypoglycaemia Fear Survey and the EQ-5D. Each severe hypoglycaemic event avoided was associated with a change of 5.9 on the Hypoglycaemia Fear Survey (HFS). Given a further estimate that each unit change on the HFS was associated with an EQ-5D quality of life change of 0.008 this led to an estimated benefit from reduced fear of severe hypoglycaemic events of 0.047 per annual event avoided. This was coupled with a direct

36

utility loss associated with a severe hypoglycaemic event in T1DM of 0.00118 to yield an overall patient benefit of 0.05 per unit reduction in annual severe hypoglycaemic events. Currie et al also reported direct disutilities in type 1 diabetes of 0.0036 per NSH event. Conclusions on hypoglycaemia Hypoglycaemia remains a major problem in type 1 diabetes and has not improved over recent Superseded – see decades. This may be because the increased emphasis on improving glycaemic control, through more intensive insulin treatment, has offset other advances in treatment; tightly managed diabetes can make it more likely that hypoglycaemia might occur. The frequency and severity of hypos can be reduced by structured education and by the use of CSII (insulin updated external pumps) but they remain a problem leading to economic disutilities. For individual events, disutilities and costs are much greater for severe hypos but the much larger number of NSHEs lead to significant impacts on quality of life. assessment report

2.2 Current service provision 2.2.1 Management of disease (15 November 2022)

In people without type 1 diabetes, the pancreas produces a little insulin throughout the day but peaks of insulin release after meals. The release after meals is very fast and enables the body to handle and store nutrients. The pancreas releases insulin into the portal vein that goes into the liver, its main site of action.

Treatment with insulin is aimed at replicating the function of the pancreas. Insulin is injected under the skin – subcutaneously. Modern insulin regimens have two components – shortacting insulin to cover mealtimes, and long-acting insulin to cover the rest of the day, usually given twice a day. The long-acting form is called basal, and the combination is often referred to as “basal-bolus” insulin, or as MDI – multiple daily injections – with three injections of short-acting insulins and two of long-acting (glargine or detemir). However, subcutaneous insulin injections cannot achieve as rapid an effect as pancreatic insulin, and because of the slower onset of action and more prolonged effects, hyperglycaemia is common shortly after meals, often followed by later hypoglycaemia.

Good control of plasma glucose by intensified insulin therapy requires more than just insulin injections. It also requires regular monitoring of blood glucose by finger-pricking and measurement using a portable meter, or by using a continuous blood glucose measurement

37

(CGM) device, and then adjustment of insulin dose to take account of calorie intake from

food and energy expenditure in exercise. People with diabetes almost always manage their

own diabetes, supported by structured education packages such as DAFNE (Dose Adjustment

for Normal Eating). The aim of treatment is to control hyperglycaemia and avoid hypoglycaemia. Glycaemic Superseded – see control is assessed using glycated haemoglobin, HbA1c, which gives an average measure over 2-3 months. The NICE target for type 1 diabetes is 48 mmol/mol (formerly 6.5%) but few people with T1DM achieve that. With the spread of continuous glucose measurement (CGM) devices, “time in range” is increasingly used as another measure of glycaemic updated external control. The alternative to MDI is continuous subcutaneous insulin infusion (CSII) using an insulin assessment report pump. CSII was approved by NICE with restrictions (see Box 1).[6]

Box 1. NICE guidance: Continuous subcutaneous insulin infusion for the treatment of diabetes mellitus [TA151] (15 November 2022)

Continuous subcutaneous insulin infusion (CSII or 'insulin pump') therapy is recommended as a treatment option for adults and children 12 years and older with type 1 diabetes mellitus provided that:

• attempts to achieve target haemoglobin A1c (HbA1c) levels with multiple daily injections (MDIs) result in the person experiencing disabling hypoglycaemia. For the purpose of this guidance, disabling hypoglycaemia is defined as the repeated and unpredictable occurrence of hypoglycaemia that results in persistent anxiety about recurrence and is associated with a significant adverse effect on quality of life or

• • HbA1c levels have remained high (that is, at 8.5% [69 mmol/mol] or above) on MDI therapy (including, if appropriate, the use of long-acting insulin analogues) despite a high level of care.

38

CSII therapy is recommended as a treatment option for children younger than 12 years with type 1 diabetes mellitus provided that:

• MDI therapy is considered to be impractical or inappropriate, and

• • children on insulin pumps would be expected to undergo a trial of MDI therapy

• Superseded – see between the ages of 12 and 18 years. updated external

• The guidance on the use of the Veo pump also had restrictions (see Box 2).[5] Box 2: NICE guidance: Integrated sensor-augmented pump therapy systems for managing blood glucose levels in type 1 diabetes (the MiniMed Paradigm Veo system and the Vibe and

• assessment report G4 PLATINUM CGM system) [DG21] 1. The MiniMed Paradigm Veo system is recommended as an option for managing

• 15 November 2022 blood glucose levels in people with type 1 diabetes only if: • they have episodes of

• ( ) disabling hypoglycaemia despite optimal management with continuous subcutaneous insulin infusion,

  2. The MiniMed Paradigm Veo system should be used under the supervision of a trained multidisciplinary team who are experienced in continuous subcutaneous insulin infusion and continuous glucose monitoring for managing type 1 diabetes only if the person or their carer: • agrees to use the sensors for at least 70% of the time • understands how to use it and is physically able to use the system and • agrees to use the system while having a structured education programme on diet and lifestyle, and counselling. 
  

  3. People who start to use the MiniMed Paradigm Veo system should only continue to use it if they have a decrease in the number of hypoglycaemic episodes that is sustained. Appropriate targets for such improvements should be set.

The guidance did not comment on reduction of severity of hypos.

39

In non-diabetic people, hypoglycaemia is rare, because if the blood glucose drops, a counterregulatory mechanism kicks in, including release of glucagon (which raises blood glucose) and adrenaline, and cessation of insulin release. In people on MDI, there are pools of long-acting and short-acting insulin under the skin (subcutaneous) which unlike pancreatic insulin, cannot be switched off. In people on CSII, there is only a little short-acting insulin, so stopping the Superseded – see pump gives a quick response. (There can be a hazard here, in that should a pump fail, the patient soon has no insulin and is at risk of hyperglycaemia and diabetic ketoacedocis (DKA). Interventions to reduce hypoglycaemia updated external One intervention to reduce the risk of hypoglycaemia is structured education such as the DAFNE Programme. Structured education is recommended in NG17 ( Recommendations | Type 1 diabetes in adults: diagnosis and management | Guidance | NICE). The assessment assessment repor report for the original appraisal of patient education in diabetes has been published in the HTA t Monograph series (Loveman et al 2003) Iqbal and Heller[24] provide a recent review of the role of structured education and (15 November 2022) hypoglycaemia. They note that until recently, the frequency of severe hypoglycaemia had not fallen over the last 20 years despite advances in treatment. They conclude that structured education can reduce the incidence of severe hypoglycaemia by about 50%, and that there is some evidence, albeit from an observational study with no control group, that the DAFNEHypoglycaemia Awareness Restoration Training (DAFNE-HART) programme can reduce hypoglycaemia even in patients with hypoglycaemia unawareness.

Continuous glucose monitoring

There are various forms of CGM. The term “continuous” is slightly misleading – glucose levels are measured every few minutes. The device measures the level of glucose under the skin (“interstitial glucose”) which reflects the level in the blood, but with a slight delay.

There are three elements in CGM

• A sensor that sits just underneath the skin and measures glucose levels.

• A transmitter attached to the sensor and sends the results to a display device.

• A display device that shows the glucose level.

40

The diabetic person checks the CGM data and adjusts insulin dose, calorie intake or activity levels to maintain blood glucose levels.

So, the traditional “loop” involves CGM, the patient using the data, and insulin dosage. Autosuspend pumps Superseded – see The mechanism here is that the CGM – patient – pump loop is augmented by direct

communication between CGM device and the pump. If blood glucose is falling too low, the CGM device communicates with the pump and switches off the insulin infusions, for say 2 hours. This is particularly useful in nocturnal hypoglycaemia when the patient is asleep. updated external Closed loop systems This term refers to systems with three components – CGM, a microprocessor with algorithms, assessment report and a pump. In effect, the microprocessor replaces the person. The microprocessor (in effect a small computer) receives data from the CGM and adjusts the infusion rate from the pump. Devices such as the Veo only control the pump when hypoglycaemia is occurring. They may (15 November 2022) switch off the insulin infusion when blood glucose falls to low, or if it is heading in that direction.

Closed loop systems can also control insulin infusion if blood glucose is too high. The most advanced system is the iLet from BetaBionics which is a dual pump which infuses insulin if blood glucose is too high, and glucagon if it is too low.

2.2.2 Variation in services and/or uncertainty about best practice

At diagnosis, the diabetes professional team should work with adults with type 1 diabetes to develop a plan for early care. Individual care plans include diabetes education, including dietary advice, insulin therapy, (including dosage adjustment, self-monitoring, avoiding hypoglycaemia and maintaining hypoglycaemia awareness), family planning, cardiovascular risk factor monitoring and management, complications monitoring and management, and communicating with the diabetes professional team. There are different factors that should be taken into account to offer an appropriate glucose monitoring device for any person. Based on individual preferences, needs, characteristics, and the functionality of the devices available, adults with type 1 diabetes may be offered a choice of glucose monitoring. Modes include real-time continuous glucose monitoring (rtCGM) or intermittently scanned

41

continuous glucose monitoring (isCGM, commonly referred to as 'flash'), these measurement

systems are coupled with multiple daily injection basal–bolus insulin regimens, or insulin pumps (Continuous subcutaneous insulin infusion (CSII) therapy), using Rapid-acting

insulin, and/or Mixed insulin.[2] People with type 1 diabetes may experience significant improvements in their lives as a result Superseded – see of the rapidly evolving technologies such as closed loop systems and artificial pancreas.[25]

Demand for these technologies is increasing, with many people with type 1 diabetes anticipated to benefit from an artificial pancreas or closed loop system in the future.[25] There is evidence using key outcomes, such as HbA1c, time in range and severe or nocturnal updated external hypoglycaemia, to demonstrate whether devices provide clinical benefits over standard selfmonitoring of blood glucose. However, quality or sample size of the studies is frequently not good enough to clearly show the clinical benefits of one technology over another. assessment report 2.2.3 Relevant national guidelines, including National Service Frameworks NICE guideline [NG17] covers care and treatment for adults (aged 18 and over) with type 1 (15 November 2022) diabetes, including advice on diagnosis, education and support, blood glucose management,

cardiovascular risk, and identifying and managing long-term complications.[2] Evidence reviews by NICE evaluated the most effective method of glucose monitoring to improve glycaemic control in adults with type 1 diabetes. Overall, 17 studies were included in clinical effectiveness analysis to examine rtCGM vs isCGM , rtCGM vs standard self-monitoring of blood glucose (SMBG), and isCGM vs SMBG. Two UK studies among 14 primary studies that contained cost utility analyses were included in this evidence review. Results show time in range (TIR) to be a better measure than HbA1c as it captures variation and can be more directly linked to risk of complications. There was a clinically meaningful positive effect on time in range for rtCGM vs both isCGM and SMBG, as well as is CGM vs SMBG, on the pre-set minimally important difference (MID) of a 5% change.[26] The authors clarified that the service user should consult with a member of the diabetes care team with expertise in the use of CGM. This guideline reported both published UK cost-effectiveness studies (one on rtCGM and one on isCGM) found these technologies to be cost-effective compared to intermittent capillary blood glucose monitoring. Based on the results of economic modelling (using clinical data from the RCTs included in the clinical review), isCGM glucose monitoring was clearly cost-effective for the overall population of people with type 1 diabetes, and this finding was robust to all the sensitivity analyses undertaken.[26]

42

The Scottish Health Technology Group (SHTG) review examined the cost-effectiveness of using closed loop systems and the artificial pancreas for the management of type 1 diabetes compared with current diabetes management options, and considered clinical effectiveness,

safety and patient aspects.[25] The evidence reviewed on the clinical effectiveness consisted of small cross-over RCTs that Superseded – see tested the use of closed loop systems over relatively short periods of time, in people with well

controlled diabetes who had had the condition for several years and who often had experience with using insulin pumps. The results of an NMA and three pairwise meta-analyses show significant improvements in mean percentage time in range for people with type 1 diabetes updated external using a closed loop system compared with other insulin-based therapies. The pairwise metaanalyses also reported statistically significant reductions in mean percentage time spent in hyperglycaemia and hypoglycaemia. High heterogeneity was present in all meta-analyses, for assessment report all outcomes. This is potentially a result of small study size, multiple different closed loops systems in the intervention group, and use of a variety of methods of insulin therapy in the control groups. It should be noted that some of the secondary evidence reviewed may be (15 November 2022) based on technologies that have since been superseded by newer models because of the rapidly changing nature of these systems.

Also, adverse events were rarely reported in either the closed loop system or control groups.

The SHTG economic model, showed that closed loop systems were associated with the highest costs and QALYs in a Scottish adult population with type 1 diabetes, except in the comparison with CGM plus CSII. Base case results showed that the technology is costeffective compared with CGM plus CSII, but not cost-effective in comparison with flash or continuous glucose monitoring combined with multiple daily injections in people with well controlled type 1 diabetes. There are some uncertainties because of a lack of published studies underpinning assumptions in the model.

2.3 Description of technology under assessment

2.3.1 Summary of Intervention

The intervention of interest is a class of automated insulin delivery systems called hybrid closed loop systems which consist of three components – a CGM, a microprocessor with control algorithms, and a pump. The microprocessor receives data from the CGM and adjusts the infusion rate from the pump, to help keep glucose levels in a healthy range. These systems are

43

aimed at reducing user or caregiver input in insulin dosing and some only require users to deliver meal boluses by entering the estimated amount of carbohydrates for meals at the time they are eaten.

There are several hybrid closed loop systems available in the UK. Some of these systems have received regulatory approval for a fixed combination of CGM, control algorithm, and insulin Superseded – see pump. However, some systems involve combining interoperable devices. The following systems are representative of the intervention of interest and have been identified by NICE as currently available in the UK. updated external Advanced HCL HCL systems use control algorithms to automate basal insulin delivery based on glucose sensor values, in order to increase the time that a patient spends in the target range and thus assessment report reduce the frequency and duration of hypoglycaemia. The user of the HCL system is required to enter their carbohydrate intake before each meal, so that the appropriate meal-time insulin (15 November 2022) bolus can be delivered by the system.

Advanced HCL (AHCL) systems have additional features that include automated correction

of bolus insulin delivered up to every 5 minutes when glucose levels are elevated. These

systems may also enable greater personalisation of insulin delivery and monitoring and can include meal detection modules that allow the system to deliver more aggressive auto correction boluses.[27]

2.3.1.1 MiniMed 670G

MiniMed 670G (Medtronic) is a CE marked hybrid closed loop system that uses a control algorithm called SmartGuard. SmartGuard technology has a manual mode and an auto mode. In manual mode, the 670G works just like other sensor-augmented pump systems. In automode function, blood glucose data measured by the CGM (Guardian sensor) is sent wirelessly to the insulin pump (670G), to enable adjustment of basal insulin every five minutes to maintain sensor glucose levels near a target glucose of 120 mg/dL (6.7 mmol/L). The system requires some user interaction to administer mealtime bolus doses. The 670G is not licensed for use in children under 7 years old. The device is also not to be used in people who require less than a

44

total daily insulin dose of 8 units per day because the device requires a minimum of 8 units per day to operate safely.

2.3.1.2 MiniMed 780G

MiniMed 780G (Medtronic) is a CE marked hybrid closed loop system launched in 2020. It has an advancement on the algorithm used in the 670G system and has Bluetooth connectivity. Superseded – see The system includes different glucose targets, according to the users’ needs. In addition to the target glucose of 120 mg/dL (6.7 mmol/L), users can also select to achieve a tighter glucose target of 5.5 - 6.1 millimoles per litre. In contrast to its predecessor system, the 780G has an updated external ‘autocorrection feature’ that delivers correction boluses automatically when sustained hyperglycemia is detected. This requires minimal user or carer interaction. The CGM (Guardian sensor) is connected to the MiniMed mobile app via Bluetooth, which optionally assessment report automatically uploads data to the CareLink connect system to notify carers or for clinician review. The 780G is not licensed for use in children under 7 years or for people who require less than a total daily insulin dose of 8 units per day because the device requires a minimum of 8 units per day to operate safely. (15 November 2022)

2.3.1.3 Control IQ

The Control-IQ (Tandem Diabetes Care) is a CE marked system that combines t:slimX2 insulin pump and Control-IQ technology. This system can be interlinked with a compatible CGM to form a hybrid closed loop system which suspends insulin delivery in response to predicted hypoglycaemia, or gives a correction bolus in response to predicted hyperglycaemia. ControlIQ has 6 settings, including optional settings for sleep and exercise, to adjust basal insulin delivery depending on user need. Mealtime bolus doses are administered manually. Data from Control-IQ can be uploaded on the Diasend or Tidepool data clouds for clinician review. Control-IQ is not licensed for use in children under 6 years or for people who require less than a total daily insulin dose of 10 units per day or who weigh less than 55 pounds, as those are the required minimum values needed to operate safely.

2.3.1.4 CamAPS FX

CamAPS FX (Camdiab) is a CE marked android app developed at the University of Cambridge. The app can be interlinked with a compatible CGM (Dexcom G6) and insulin pump (Dana RS or Dana-I) to form a hybrid closed loop system. CamAPS FX can operate on an auto mode ‘off’ whereby basal insulin delivery is pre-programmed by the user or an auto mode ‘on’ where

45

insulin delivery is directed by the app. In auto mode on, a bolus dose calculator embedded in the app allows the user to initiate the delivery of mealtime insulin dose. If the auto mode ‘on’ feature is prevented from coming on, an auto mode ‘attempting’ feature is initiated in which insulin delivery is reverted to pre-programmed basal rates. Data from CamAPS FX can be uploaded to the Diasend data cloud, for clinician review. CamAPS FX is licensed for use in Superseded – see people aged 1 year and older and in pregnancy, however, other age restrictions may apply depending on the chosen CGM and insulin pump. 2.3.2 Identification of important sub-groups updated external The NICE scope (March 2022) states the following subgroups if evidence permits: o Women with type 1 diabetes who are pregnant and those planning pregnancy (not including gestational diabetes). Note that in this assessment this subpopulation is not assessment report required to fulfil the criteria of prior use of at least 1 technology. o Children with type 1 diabetes. o If possible, evidence should be analysed based on the following age groups: o 5 years and under, (15 November 2022)

o 6 - 11 years

o 12 -19 years

• People with extreme fear of hypoglycaemia

• People with diabetes related complications that are at risk of deterioration

2.3.3 Current usage in the NHS

The management of T1DM involves lifestyle adjustments, monitoring of blood glucose levels, and insulin replacement therapy, with the aim of recreating normal fluctuations in circulating insulin concentrations. Blood glucose levels are monitored to determine the type and amount of insulin needed to regulate blood glucose levels and reduce the risk of complications.

NICE guidelines recommend that adult and pregnant women with T1DM should be empowered to self-monitor their blood glucose, supported by structured education packages (e.g., Dose Adjustment for Normal Eating) on how to measure glucose levels and interpret the results.[2] NICE also recommends that children and young people with T1DM and their families or carers should be offered a continuing programme of education from diagnosis. Several systems of monitoring glucose levels and delivering insulin are available in clinical practice. The system

46

recommended for individuals is based on the individual’s age, whether they are pregnant, their glycaemic control, and personal preferences (Figure 1).

Superseded – see updated external assessment report (15 November 2022)

47

Superseded – see updated external assessment report (15 November 2022)

Figure 1. Management of type 1 diabetes mellitus (www.nice.org.uk/guidance/ng17)

48

2.3.3.1 Blood glucose monitoring

Capillary blood glucose monitoring

Blood glucose concentrations in diabetes can vary considerable from day-to-day and over the course of a 24‑hour period. Routine blood glucose testing is typically done using capillary blood glucose monitoring. Capillary blood glucose monitoring involves pricking a part of the Superseded – see body (usually the finger) with a lancet device to obtain a small blood sample at certain times of the day. The drop of blood is then applied to a test strip which is inserted into a blood glucose meter for automated determination of the glucose concentration in the blood sample at the time updated external of the test. Blood glucose measurements are taken after several hours of fasting, usually in the

morning before breakfast, and before and after each meal to measure the change in glucose concentration. assessment report

• NICE recommends routine self-monitoring of blood glucose levels at fingertips for all adults with T1DM at least 4 times a day, including before each meal and before bed.[2] For pregnant

• (15 November 2022) women with T1DM, the NICE recommendation is to test fasting, pre-meal, 1-hour post-meal, and bedtime blood glucose levels daily. The NICE recommendation for children and young people with T1DM is capillary blood glucose testing 5 times per day.[28]

Real time continuous blood glucose measurement (rtCGM)

rtCGM is an alternative to routine finger-prick blood glucose monitoring for people (including pregnant women) aged 2 and over, who have diabetes, have multiple daily injections of insulin or use insulin pumps, and are self-managing their diabetes. This involves measuring interstitial fluid glucose levels throughout the day and night.

A rtCGM system comprises three parts:

• A sensor that sits just underneath the skin and measures glucose levels

• A transmitter that is attached to the sensor and sends glucose levels to a display device

• A display device that shows the glucose level (separate handheld device (known as “standalone” CGM) or a pump (known as an “integrated system”)

49

For most rtCGM systems, calibration by checking the finger-prick blood glucose level is needed once or twice a day. rtCGM systems monitors glucose levels regularly (approximately every 5 minutes), and alerts can be set for high, low or rate of change.

NICE does not recommend offering rtCGM routinely to adults with T1DM. Instead, rtCGM Superseded – see with an alarm should be considered for adults with T1DM for whom standard management of blood glucose levels has not worked or been difficult, i.e., those with recurrent severe hypoglycaemia or impaired awareness of hypoglycaemia. The users must also be willing to commit to using the technology at least 70% of the time and to calibrate it as needed. For updated external

children and young people with T1DM, NICE recommends that ongoing rtCGM with alarms should be offered to those who continue to have severe hypoglycaemia or impaired hypoglycaemia awareness, or those who are not able to recognise or communicate symptoms assessment report of hypoglycaemia. The NICE recommendation is to offer rtCGM to all pregnant women with T1DM to help them meet their pregnancy blood glucose targets and improve neonatal outcomes. (15 November 2022)

Flash/intermittently scanned glucose monitoring

Flash glucose monitoring systems comprise a reader and a sensor applied to the skin to measure interstitial fluid glucose levels. It only provides a reading or trends when the sensor is scanned. The NICE guidelines for adults and children with T1DM do not comment on the use of flash systems for intermittent interstitial fluid glucose monitoring.

For pregnant women with T1DM, the NICE recommendation is to offer intermittently scanned flash monitoring to those who are unable to use rtCGM or express a clear preference for it. In standard practice and in accordance with the NHS long-term plan, most centres offer flash and/or CGM to pregnant women with T1DM.

HbA1c

Longer-term control is measured by glycated haemoglobin levels (HbA1c), which reflect the average blood glucose levels over 2 to 3 months. HbA1c is correlated to CGM results over the preceding 8-to-12 weeks.[29] NICE guidelines on diabetes (type 1 and type 2) in children and young people, adults, and diabetes in pregnancy recommend that people with T1DM should

50

aim for a target HbA1c level of 6.5% (48 mmol/mol) or lower to minimise the risk of long term complications from diabetes. Poor glycaemic control may trigger a discussion about different options for insulin administration.

2.3.3.2 Insulin regimens Supers Multiple daily injections (MDI) eded – see

Insulin is injected subcutaneously. Modern insulin regimens have two components – shortacting insulin to cover mealtimes, and long-acting insulin to cover the rest of the day, which is usually given twice a day. The long-acting form is called basal, and the combination is often updated external referred to as “basal-bolus” insulin, or as multiple daily injections (MDI), with three injections of short-acting insulins and one or two of long-acting insulin. However, subcutaneous insulin injections cannot achieve as rapid an effect as pancreatic insulin, and because of the slower assessment report onset of action and more prolonged effect, hyperglycaemia is common shortly after meals, (15 November 2022) often followed by hypoglycaemia later.

The NICE recommendation is to offer MDI basal–bolus insulin regimens for all adults, children

and young people with T1DM. For pregnant women with diabetes, NICE recommends that rapid-acting insulin analogues should be considered.

Continuous subcutaneous insulin infusion (CSII)

The alternative to MDI is continuous subcutaneous insulin infusion (CSII) using an insulin pump. It makes use of an external pump that delivers insulin continuously from a refillable storage reservoir by means of a subcutaneously placed cannula. CSII was approved by NICE as a treatment option for adults and children 12 years and older with T1DM provided that:

• attempts to achieve target HbA1c levels with MDIs result in the person experiencing disabling hypoglycaemia. For the purpose of this guidance, disabling hypoglycaemia is defined as the repeated and unpredictable occurrence of hypoglycaemia that results in persistent anxiety about recurrence and is associated with a significant adverse effect on quality of life, or

• HbA1c levels have remained high (that is, at 8.5% (69 mmol/mol) or above) on MDI therapy (including, if appropriate, the use of long-acting insulin analogues) despite a high level of care.

51

CSII therapy is recommended as a treatment option for children younger than 12 years with T1DM provided that:

• MDI therapy is considered to be impractical or inappropriate, and • children on insulin pumps would be expected to undergo a trial of MDI therapy Superseded – see between the ages of 12 and 18 years. For pregnant women with T1DM, NICE recommends that CSII should be offered to women who are using MDI and do not achieve blood glucose control without significant disabling updated external

hypoglycaemia. - Integrated sensor augmented pump therapy systems (SAP) assessmen Integrated sensor-augmented pump therapy systems combine rtCGM with CSII. The systems t report

are designed to measure interstitial glucose levels (every few minutes) and allow immediate real‑time adjustment of insulin therapy. The systems may produce alerts if the glucose levels become too high or too low. NICE’s diagnostic guidance (DG21) on integrated sensor(15 November 2022)

augmented pump therapy systems for managing blood glucose levels in T1DM recommends the MiniMed Paradigm Veo system as an option for managing blood glucose levels in people with T1DM only if they have episodes of disabling hypoglycaemia despite optimal management with CSII.[5] As with other pumps the user can program one or more basal rate settings for different times of the day/night. A built-in bolus calculator works out how much insulin is needed for a meal following the input of carbohydrates consumed. The advanced feature of sensor-augmented pump is that the rtCGM – patient – pump loop is augmented by direct communication between the rtCGM device and the pump. If blood glucose is falling too low, the rtCGM device communicates with the pump and automatically switches off (suspends) the insulin infusions. Depending on the device, the user either must restart insulin delivery or the pump resumes insulin delivery after 2 hours.

LGS/PLGS

SAP systems can operate in standard (manual) and advanced (automatic) modes. In the manual open loop mode, the continuous glucose monitor and glucose pump do not communicate with each other, and insulin doses are programmed by the user, who makes manual adjustments.

52

In advanced, automatic mode, the CGM device and pump can communicate with each other automatically based on real-time glucose data, in order to adjust the insulin basal rate and suspend the insulin infusion without the input of the wearer in order to prevent potential

hypoglycaemia. Glucose suspension can be a simple ‘low glucose suspend’ (LGS) function, in which insulin infusion is suspended when glucose monitoring systems detect that glucose Superseded – see levels have fallen below a specific hypoglycaemia threshold. In this case, insulin is

suspended for a period of time and may resume when the system determines that glucose levels have returned to within target range or when the glucose suspension is overridden by the patient. updated external Predictive low glucose suspend (PLGS) is a more advanced use of technology in which prediction algorithms are used which essentially forecast future hypoglycaemia (e.g. within the assessment report

next half hour), and pre-emptively suspend insulin delivery before hypoglycaemia develops. PLGS systems will then automatically resume insulin infusions if the user overrides the (15 November 2022) suspension, or if glucose levels begin to rise or rise above a specific threshold.[30, 31]

3 DEFINITION OF THE DECISION PROBLEM

3.1 Decision problem

3.1.1 Interventions

The interventions of interest are hybrid closed loop systems - a class of automated insulin delivery systems which consists of three components – a CGM, a microprocessor with control algorithms, and a pump.

There are several hybrid closed loop systems available in the UK such as MiniMed 670G and MiniMed 780G. The systems are representative of the intervention of interest and have been identified by NICE as currently available in the UK.

53

Population and sub-groups are per NICE scope (published March 2022).

3.1.2 Population including sub-groups

Populations People who have T1DM who are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring[ab ] Superseded – see If evidence permits the following T1DM subpopulations will be included: • Pregnant women and those planning pregnancies (excluding gestational updated external diabetes).[b] • Children (5 years and under, 6 – 11 years, 12 - 19 years). • People with extreme fear of hypoglycaemia. assessment report • People with diabetes related complications that are at risk of deterioration. a For the purpose of this review, difficulty refers to (1) not maintaining HbA1c levels of 6.5% (48 mmol/mol) or below (for pregnant women/those (15 November 2022) planning pregnancies: not maintaining fasting plasma glucose levels of 5.2 mmol/l or below, or not maintaining non-fasting plasma glucose of 7.7 mmol/L (one hour after eating)/ 6.3 mmol/L (two hours after eating)), (2) not maintaining at least 70% time in range of 3.9 -10 mmol/l, or (3) repeated hypoglycaemia that causes anxiety about recurrence and is associated with a significant adverse effect on quality of life. b Pregnant women and those planning pregnancies will not be required to have previously used CSII and self-monitoring of blood glucose or glucose monitoring (rt-CGM/flash glucose monitoring) with multiple daily injections. 3.1.3 Relevant comparators • Real time continuous glucose monitoring with continuous subcutaneous Comparator insulin infusion (non-integrated). • Intermittently scanned (flash) glucose monitoring with continuous subcutaneous insulin infusion.

54

55

==> picture [569 x 330] intentionally omitted <==

----- Start of picture text -----
• Miscarriage related to fetal abnormality
• Increased proportion of babies delivered by caesarean section
• Macrosomia (excessive birth weight)
• Respiratory distress syndrome in the new-born
Device related outcomes –
• Adverse events related to the use of devices se e
Superseded
Patient-reported outcomes
• Heath-related quality of life
• Psychological well being
updated external • Impact on patient (time spent managing the condition, time spent off work or school, ability
to participate in daily life, time spent at clinics, impact on sleep)
• Anxiety about experiencing hypoglycaemia
• Acceptability of testing and method of insulin administration rt
assessment repo
Carer reported outcomes
Impact on carer (fear of hypoglycaemia, time spent managing the condition, time spent off work,
ability to participate in daily life, time spent at clinics, impact on sleep)
(15 November 2022)
----- End of picture text -----

3.2 Overall aims and objectives of assessment

The overall objectives of this project are to examine the clinical and cost-effectiveness of hybrid closed loop systems for managing glucose levels in people who have T1DM. The key questions for this review are provided in the box below.

Key question 1

What is the clinical effectiveness of hybrid closed loop systems for managing glucose in people who have T1DM and are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring?

Sub questions

1. What is the clinical effectiveness of hybrid closed loop systems for managing glucose in pregnant women who have T1DM?

56

57

2. What is the cost effectiveness of hybrid closed loop systems for managing glucose in

children who have T1DM and are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring? Superseded – see

3. What is the cost effectiveness of hybrid closed loop systems for managing glucose in people who have T1DM, an extreme fear of hypoglycaemia, and are having difficulty managing their condition despite prior use of at least one of the following

updated external technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring?

assessment report

4. What is the cost effectiveness of hybrid closed loop systems for managing glucose in people who have T1DM, with diabetes related comorbidities that are at risk of (15 November 2022) deterioration, and are having difficulty managing their condition despite prior use of at least one of the following technologies: continuous subcutaneous insulin infusion, real time continuous glucose monitoring, flash glucose monitoring?

58

4 ASSESSMENT OF CLINICAL EFFECTIVENESS

Systematic review methods followed the principles outlined in the Cochrane Handbook of

Diagnostic Test Accuracy[32] and the NICE Diagnostic Assessment Programme manual.[33]

4.1 Methods for reviewing effectiveness Superseded – see 4.1.1 Identification of studies 4.1.2 Search strategy The search strategy comprised the following main elements: updated external

1. Searching of electronic bibliographic databases and other online sources,

2. Searching of electronic bibliographic databases and other online sources, 2) Contacting experts in the field, and assessment report 3) Scrutiny of references of included studies, relevant systematic reviews, and the most recent NICE guidance on systems that combine CGM and CSII.[5] A comprehensive search was developed iteratively and undertaken in a range of relevant (15 November 2022) bibliographic databases and other sources, following the recommendations in Chapter 4 of the

Cochrane Handbook for Systematic Reviews of Interventions.[34] Search terms were related to T1DM (including a separate set of terms relating to pregnant women and women planning pregnancy) and technologies to manage blood glucose levels. Search strings applied in the previous technology assessment on integrated sensor-augmented pump therapy systems (DG21)[35] were used as the basis for developing selected lines relating to type 1 diabetes, insulin pumps, sensor augmented pumps and multiple daily injections, and other systematic reviews informed the lines relating to pregnancy.[36-38] The main MEDLINE search strategies were independently peer reviewed by a second Information Specialist.

Date limits were used, in order to identify records added to databases since the searches for DG21 (run in 2014).[35] Searches were conducted in March and April 2021, and updated in April 2022, in the following resources: MEDLINE ALL (Ovid); Embase (Ovid); Science Citation Index and Conference Proceedings (Web of Science); Cochrane Database of Systematic Reviews (Wiley); CENTRAL (Wiley); Clinicaltrials.gov; HTA database (CRD); International HTA database (INAHTA); NIHR Journals Library; and the following websites:

• U.S. Food & Drug Administration (FDA)

59

• Medicines & Healthcare Products Regulatory Agency (MHRA)

• Agency for Healthcare Research and Quality (AHRQ)

• Canadian Agency for Drugs and Technologies in Health (CADTH)

• Swedish Agency For Health Technology Assessment And Assessment Of Social Services (SBU) Superseded – see The search was developed in MEDLINE (Ovid) and adapted as appropriate for other resources. Full search strategies are provided in Appendix 1: Record of searches – Clinical effectiveness (see section 10.1.1). updated external

Records were exported to EndNote X9, where duplicates were systematically identified and

removed. Where available, alerts were set up so that the team were aware of any new, relevant publications added to databases beyond the original search date. assessment report 4.1.3 Inclusion and exclusion criteria Studies that satisfy the following criteria were included: (15 November 2022)

60

61

62

• Impact on carer (fear of hypoglycaemia, time spent managing the condition, time spent off work, ability to participate in daily life, time spent at clinics, impact on sleep) Study design Hybrid closed loop systems studies Su persed • Any design ed – see All comparator studies • Comparative effectiveness study designs updated external Healthcare Self-use supervised by primary or secondary care setting t assessment repor Publication Peer reviewed papers type Abstracts and manufacturer data will be included only if they provide numerical (15 November 2022) data and sufficient detail on methodology to enable assessment of study quality/risk of bias. Further, only data on outcomes that have not been reported in peer-reviewed full text papers will be extracted and reported. Language English

Research papers were included where it could not be established if all study participants had difficulty managing their condition (defined by HbA1c, fasting plasma glucose, non-fasting plasma glucose, or time in range as above), if the group mean met this criterion.

Papers that fulfilled the following criteria have been excluded:

Non-human studies, letters, editorials, and communications. Qualitative studies. Studies conducted outside of routine clinical care settings, e.g., inpatient research facilities, diabetic summer camps. Studies where more than 10% of the sample did not meet the inclusion criteria (for example over 10% were inpatients). Studies without extractable numerical data. Studies that provided insufficient information for assessment of methodological quality/risk of bias. Articles not available in the English language. Studies evaluating individual components and

63

not complete hybrid close loop systems. Studies of DIY closed loop systems, which are not approved by regulatory bodies.[39] Studies evaluating automated insulin delivery systems which only suspend insulin delivery when glucose levels are low/ are predicted to get low. 4.1.4 Review strategy Superseded – see 4.1.4.1 Prioritization strategy for full text assessment We applied a two-step approach for identifying and assessing relevant evidence. We applied stricter criteria at the point of data extraction/risk of bias than title and abstract assessment to updated external prioritise and select the best available evidence.[40-42] The elements used to prioritise evidence

(study design, study length, sample size) were chosen in collaboration with NICE and diabetes clinicians as those that will provide the most applicable evidence. assessment report Step one: The studies were scoped in Endnote before deciding which studies qualified for full text assessment (step two). Records were coded in terms of study design and study duration. Randomised controlled trials (RCTs) were prioritised over controlled trials. Non-randomised (15 November 2022) controlled trials/comparative effectiveness studies were prioritised over non-comparative studies. Longer term studies (6 months or more) were prioritised (see section 4.1.4.1) over shorter-term studies.

Step two: studies identified from step one went through the standard systematic reviewing approach of full text assessment. We followed the pre-defined PICO (see for study 4.1.3 eligibility criteria) to assess the eligibility of studies.

4.1.4.2 Prioritization strategy for data extraction and risk of bias Given the limited time and resources available, deprioritised studies i.e. the large number of observational studies which otherwise met the inclusion criteria for this review were narratively reported and listed. RCTs were prioritised for data extraction and quality assessment.[42] .

4.1.5 Data abstraction strategy

We extracted the following study characteristics:

64

Details on study design and methodology, participant characteristics, intervention characteristics, comparator characteristics, outcomes, outcome measures, and additional notes (such as funding).

Two reviewers extracted data independently, using a piloted data extraction form. Disagreements was resolved through consensus, with the inclusion of a third reviewer when Superseded – see required. 4.1.6 Critical appraisal strategy The risk of bias of randomised trials was assessed using the revised Cochrane risk-of-bias tool updated external for randomized trials (RoB 2).[43] Risk of bias in controlled trials, non-randomised trials, and cohort studies was assessed using the Cochrane risk of bias in non-randomized studies of interventions (ROBINS-I) tool.[44] Risk of bias for case control studies and controlled beforeassessment report and-after studies was assessed using Effective Practice and Organisation of Care (EPOC) RoB Tool.[45] Two reviewers assessed risks of bias. Disagreements were resolved through consensus, with the inclusion of a third reviewer if required. (15 November 2022)

4.1.7 Methods of data analysis/synthesis

We synthesised the RCT evidence statistically. The network meta-analysis was conducted using a frequentist approach and a random-effects model.

Subgroup analyses were undertaken where possible for the different combinations of interventions study participants had previously used to manage their blood glucose (i.e., flash glucose monitor and multiple daily insulin injections, flash glucose monitor and CSII, rtCGM and multiple daily insulin injections, rtCGM and CSII, self-blood glucose monitoring and CSII).

4.1.7.1 Pairwise and network meta-analysis

The analysis compared hybrid close-loop systems and relevant comparators for managing blood glucose levels in T1DM. The primary effectiveness outcome was HbA1c. Other clinically relevant outcomes include the ‘time in target range’ which gives the percentage of time that a person spends with blood glucose level in target range of 70 to 180mg/dl, and adverse events (e.g., severe hypoglycaemia, diabetic ketoacidosis).

Decisions about information to include in the NMA were informed by relevance to the decision problem and sufficient similarity across studies (e.g., patient characteristics and study design)

65

to reduce the risk of violating underlying assumptions of transitivity/coherence when pooling direct and indirect evidence across studies. We used an iterative process[46] to define the extent of the treatment network and to identify studies for inclusion. This involved first defining an initial core set of interventions that met the criteria set out in the projects’ scope and included trials of such interventions in T1DM populations. Superseded – see Publication bias was assessed visually using a comparison-adjusted funnel plot, where

• publication bias is present if the funnel plot is asymmetrical. Egger’s test was also used, where publication bias is considered to exist if p<0.05. Transitivity was assessed by looking at the distributions of potential effect modifiers across all updated external studies included in the systematic review. To check for consistency of each network, net splitting can be performed which splits the estimates in the network into direct and indirect estimates. Statistically significant

• assessment report inconsistency is present between the direct and indirect estimates if the p-value of the difference between effect estimates is <0.05. However, due to the small number of studies and treatments in each network, net splitting was not feasible. Loop consistency was also not tested

• (15 November 2022) as there were no closed loops in the networks for any of the outcomes.

Treatments were ranked using P-score, which measures the certainty that one treatment is better than another treatment, averaged over all competing treatments.

Statistical analyses were performed using RStudio version 4.1.0.

4.1.8 Dealing with missing data

We conducted the review according to the registered protocol.

4.2 Results

4.2.1.1 Number of studies identified

The literature search provided 12890 records potentially related to the area of interest; 7292 records remained after removing duplicates. After the abstract screening, 1364 records were identified for full paper screening. A further 1326 articles were excluded at the full-text stage mainly due to incorrect intervention/comparators, study design, incorrect population, abstract/poster presentation only or further duplication identified. 14 records (12 RCTs)[27, 47-] 59 and 9 observational studies 27, 60-65 are presented for this systematic review of clinical

66

effectiveness. Three papers drew on the same study participants. External submissions, including NHS England evidence and company submissions are also presented in this report. The PRISMA flow diagram is shown in the figure below.

==> picture [569 x 578] intentionally omitted <==

----- Start of picture text -----
Records identified Duplicate records
(n = 12890) removed
Superseded – see (n = 5598)
updated external
Records screened (after duplicates removed) Records excluded at
(n = 7292) title and abstract level
(n = 5928)
assessment report
(15 November 2022)
Full-text articles
excluded
(n = 1326)
Full-text articles assessed for eligibility
(n = 1364)
Full-text records included in quantitative NHS audit (n = 2)
synthesis
(n = 14)
Full-text records included in qualitative
synthesis
(n= 9)
Observational studies recorded (n=17)
----- End of picture text -----

67

4.2.1.2 Number and type of studies included

Randomised controlled trials

Randomised studies

Eleven RCTs (one with two relevant intervention arms,54 13 records) 47-57, 59 were identified that yielded data of potential relevance to the decision problem assessing HCL Superseded – see against a comparator. RCTs in which HCL treatment was received for ≥ 4 weeks (range 4 to 26 weeks) were included if the comparator was relevant to the decision problem (comparators were classified as CSII + CGM and LGS/PLGS). updated external Most of these studies reported results for outcomes relevant to monitoring glycaemic control. These data were assembled using CGM technology that accumulates large amount of data and they assessed change in % time in range over a specified period of observation (baseline assessment report to final). Most studies reported change in HbA1c level (final minus baseline values). The RCTs thus provided quantitative data potentially amenable to network meta-analysis. Two Publications (Bergenstal 2021 27 and Weinzimer 2022 58) were derived from the FLAIR (15 November 2022) study and presented data comparing different types of AHCL; since HCL has been viewed here as a generic intervention the FLAIR study can be considered more similar to a single arm study (with two subgroups) than an RCT and is considered in the section describing single arm studies.

These RCTs were heterogeneous in multiple respects including trial design (parallel groups or cross over design with wash-out phase between different treatments), participants’ age, number of participants, and other demographics including run-in times, duration of observation periods, and number and types of previous treatments. Studies screened relatively small numbers of patients. The number of participants randomised ranged from < 20 to 135. Table 1 summarises the main characteristics of patients recruited in RCTs with treatment duration 4 to 26 months (additional RCT details are in 10.2. Most studies were conducted in children or young adults. For young children it would likely be difficult to clearly establish whether they were having difficulty in controlling glycaemia prior to recruitment. Only McAuley 2022 51 and Boughton 2019 48 looked at HCL use in elderly patients (age >60 years); in control arm for practical reasons and familiarity with method the participants continued with their previous method of glycaemic control which presumably was long

68

established (i.e. they were not “re-trained” in a new non-HCL method). In treatment arm participants were trained and then transferred to HCL. Both these studies in the elderly enrolled relatively few patients.

The major outcomes reported in the RCTs related to monitoring glycaemic control.

69

These included change in % HbA1c and % time within, above or below a defined blood glucose level (mmol/ litre) including: % time within range indicating satisfactory control (3.9 to 10 mmol/litre, % time in a hyperglycaemic range ( > 10 mmol/litre), and % time in a

hypoglycaemic range variously <3.9, <3.5, <3.3, <3.0 and < 2.8 mmol/litre depending on study. Low rates of severe hypoglycaemia and of ketotic episodes were also reported; it may Superseded – see be that the small number of participants and relatively short treatment periods mean that accurate estimates of the rates of these events is difficult. The outcomes reported in RCTs are summarised in Table 2. Additional outcomes are reported in updated external

Table 2. Glycaemic-control outcomes reported in RCTs of potential relevance

Outcome results reported in the RCTs are summarised below in Table 2 and presented graphically in forest plots. Glycaemic control outcomes by study arm were reported in various ways, as mean (± sd) or median (IQR) values, often baseline values for each arm were not reported or were unclear so that change from baseline was sometimes and or

70

unreported and only end of treatment values were provided. Trials reported mean difference and 95% CI between arms whether this was derived from median or mean estimates for the outcome. These reported values were available for NMA. Where necessary some outcome results have been calculated from numerical data in the relevant published reports; these together with most other data reported, were often strongly rounded to only a few decimal Superseded – see places. Table 3 summarises the data extracted from the included RCTs. We present combined results of all RCTs together covering all subpopulations, before presenting results by individual subpopulations. updated external assessment report (15 November 2022)

71

Superseded – see updated external assessment report (15 November 2022)

72

Table 3. Summary of main outcome measure reported in RCTs

73

74

75