Protocol for the Diabetes Technology Network UK and Association of British Clinical Diabetologists’ closed-loop insulin delivery audit programme

Thomas Sj Crabtree,^1,2,3 Tomás P Griffin,^4,5Alistair Lumb,^6,7 Peter Hammond,⁸ Robert Ej Ryder,² Pratik Choudhary,^4,5 Emma G Wilmot^1,3

¹Department of Diabetes & Endocrinology, Royal Derby Hospital, University Hospitals of Derby and Burton NHS Trusts, UK
²Department of Diabetes & Endocrinology, City Hospital, Sandwell and West Birmingham Hospitals NHS Trust, UK
³School of Medicine, Faculty of Medicine and Health Sciences, University of Nottingham, UK
⁴Leicester Diabetes Centre, University Hospitals of Leicester, Leicester, UK
⁵Diabetes Research Centre, College of Health Sciences, University of Leicester, Leicester, UK
⁶Oxford Centre for Diabetes Endocrinology and Metabolism, Oxford University Hospitals NHS Trust, Oxford, UK
⁷NIHR Oxford Biomedical Research Centre, Oxford, UK
⁸ Department of Diabetes & Endocrinology, Harrogate and District NHS Trust, Harrogate, UK

Address for correspondence: Dr Emma G Wilmot Department of Diabetes & Endocrinology, Royal Derby Hospital, Uttoxeter Road, Derby DE22 3NE, UK
E-mail: Emma.Wilmot2@nhs.net

Abstract

Background: The Association of British Clinical Diabetologists (ABCD) closed-loop audit aims to capture real-world out- comes from all who use hybrid closed-loop (HCL) insulin delivery systems in routine clinical care. In addition, NHS England has announced a pilot programme this year to expand access to HCL insulin delivery systems to people with type 1 diabetes (T1D) who are already using pump therapy and FreeStyle Libre with a HbA_1c ≥ 69mmol/mol (≥ 8.5%). This group is often underrepresented in current randomised control trial evidence and, vitally, the planned audit will capture their data.

Methods: The ABCD nationwide audit programme has Caldicott guardian approval and has also been approved by Conﬁdentiality Advisory Group (CAG). Clinical teams collect anonymised user data using a secure online tool. Baseline characteristics and routinely collected outcome data at follow-up will include: assessment of glycaemic outcomes ( HbA_1c, time in range, time below range); patient-reported outcome measures (Gold score and diabetes-related distress); and frequency of resource utilisation (hospital admissions, paramedic callouts, diabetic ketoacidosis [DKA] and severe hypoglycaemia).

Discussion: The ABCD closed-loop audit will produce an inde- pendent real-world dataset of outcomes in closed-loop users across multiple systems. These data will provide insight into the real-world beneﬁts and challenges of HCL systems used within the NHS in England.

Introduction

Hybrid closed-loop (HCL) insulin delivery systems combine continuous subcutaneous insulin infusions (CSII, or insulin pump therapy) with real-time continuous glucose monitoring (rtCGM) and an algorithm. The algorithm, which is held on a smartphone or within the insulin pump, receives glucose data from the rtCGM and communicates a decision to sustain, increase, decrease or suspend insulin delivery as needed to maintain glucose within a pre-speciﬁed target range.

Randomised control trials (RCTs) demonstrate improvements in HbA_1c and time-in-range and reductions in hypoglycaemia on closed-loop therapy compared to sensor-augmented pump therapy with low glucose suspend.^1-3 HCL systems have also been associ- ated with improvements in diabetes distress and other quality of life metrics.^4,5 Though these results are encouraging, the people included in these trials tended to have HbA_1c at or close to target levels prior to commencing HCL therapy. Additionally, participants in RCTs are monitored closely for evidence of adverse events and supported at a level that may be more intensive than is generally practicable for most health services. Nonetheless, even outside RCTs HCL usage demonstrates reductions in HbA_1c and improvements in time-in-range across a range of currently available systems.^6,7 Whilst these publications feature large cohorts, they do not include people with elevated HbA_1c levels at baseline and they may include people upgrading from earlier versions of closed-loop technology (e.g. Basal-IQ to Control-IQ in the Tandem trial). Also, these real-world trials tend not to include data on important outcomes such as hospital admissions, severe hypoglycaemia and patient-reported outcomes.^6,7

Of the 218,670 people with T1D captured by the National Diabetes Audit in England and Wales, only 1 in 10 were using insulin pump therapy.⁸ Data on access to rtCGM are limited and although more than 30% of the population now use FreeStyle Libre for glucose monitoring this cannot at present be used in a commercially available HCL system.⁹ It follows that the use of HCL systems until recently was limited to a group of people with diabetes who met the funding criteria for insulin pump therapy (Box 1)¹⁰ and were able to self-fund rtCGM sensors or to meet the previously strict NHS funding criteria for rtCGM. Examples include those with recurrent severe hypoglycaemia and impaired awareness of hypoglycaemia. Real-world evidence for HCL use in a UK context is therefore lacking.

Box 1 Current criteria for NHS insulin pump funding in adults as per NICE⁸

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 (HbA_1c) 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

HbA_1c 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.

Even amongst individuals who can access pumps and rtCGM on existing criteria, some may remain on older systems where the interaction between pump and rtCGM is limited to suspending glucose in anticipation of low glucose levels – known as predictive low- glucose suspend (PLGS). An overview of HCL systems commonly encountered in UK practice at the time of writing is available for reference.¹¹

The ABCD closed-loop audit launched in July 2021, the same year in which NHS England launched their HCL pilot in adults and children with T1D.¹² In line with the published diabetes technology pathway,¹³ adults with T1D who were currently using insulin pump therapy and FreeStyle Libre with a HbA_1c ≥69mmol/mol (≥ 8.5%) were eligible to access HCL technology as part of the pilot. In addition to those included in the pilot scheme, the audit will also allow data to be collected from all existing and future HCL users, with the potential for further data collection from those who may be granted access to the systems if criteria change in future. This will include those changing from a PLGS system and those commencing HCL with other criteria such as pregnancy. There were no formal exclusion criteria in this audit.

The aim of this audit programme is to capture the routine clinical outcomes of the users of HCL systems to provide real-world insights into the safety and effectiveness of closed-loop systems.

Audit development and methods

The ABCD audit has been developed by the ABCD Diabetes Technology Network UK (DTN-UK) steering group with expertise in diabetes technology. The data to be collected were determined by the steering group, who balanced the importance of each covariate or outcome in determining the safety and efﬁcacy of the systems with the data that are likely to be routinely collected within participating diabetes clinics (and therefore available to audit). A secure online tool has been developed to collect the data.

Centres are required to register in order to access the tool, and all individuals who request access are validated before access is granted. Site type is also recorded, which may in future allow differentiation between community and acute hospital-based services. Centre lead details are stored on a secure NHS server and managed by the ABCD audit administrator. The audit has been advertised by ABCD so that any centre with closed-loop insulin system users can choose to participate. As such, the audit has the potential to capture data from a broad range of individuals, which might include people transitioning from PLGS; MDI combined with is/rtCGM sensors or those previously using a pump in isolation.

To ensure anonymization, the patient identiﬁer is encrypted and only the encrypted identiﬁer is stored by the system. Users in the submitting site can search for HCL users from their own service using the NHS number, but they can only access the audit tool from within the secure NHS computer network. Outside the submitting centre, those analysing the data only see the encrypted patient identiﬁer. Further, the date of birth is converted to age by the system and only the age is stored. Data can be collected contemporaneously and entered directly into the online tool or, if more convenient, can also be collected in an editable PDF or paper form for later upload. Data are entered by clinicians at each site, and they are responsible for ensuring the validity of the data. The paper forms are included in appendix 1&2 (online at www.bjd-abcd.com). Whilst the audit is intended to be prospective from the time of commencing HCL system use, data may also be collected retrospectively should this be required for any existing users. However, patient-reported outcome measures cannot be retrospectively recalled and therefore will only be available if documented in the medical notes at a point contemporaneous to the baseline visit date.

The NHS supports clinical audits and mandates them to collect data and outcomes to help improve the service and to evaluate the use of therapies in real-world practice.⁹ As a clinical audit, this programme only collects anonymised, routinely available clinical data. Data or tests not performed routinely were not required for this audit. As the audit comprises routinely collected healthcare data only, there is no requirement for approval by a research ethics committee.¹⁴ The ABCD nationwide audit programme, which includes this audit, has Caldicott guardian approval and has also been approved by Conﬁdentiality Advisory Group.¹⁵

A range of clinical parameters will be collected at baseline, prior to HCL initiation. The baseline date will be deﬁned as the date of HCL commencement. Baseline characteristics include age, ethnicity, diabetes type, diabetes duration and postcode-derived index of multiple deprivation decile as an assessment of socio-economic status,¹⁶ and information about the HCL system and insulin being used. Retinopathy status, including grading where available, will be recorded. The frequency of hospital admissions, paramedic callouts and severe hypoglycaemia (not resulting in paramedic response but requiring third party assistance) in the 12 months before starting the HCL system will be captured alongside HbA_1c, weight and height. Two validated, routinely used scoring systems will be utilised: Gold Score to assess hypoglycaemia awareness and the Diabetes Distress Score.^17,18 FreeStyle Libre glycaemic metrics, including time in range, time below range, time above range, glucose management indicator, scan frequency and coefﬁcient of variation, will be recorded at baseline, using ranges deﬁned by international consensus.¹⁹

The primary measure of interest is change in laboratory-derived HbA_1c. Changes in weight, body mass index (BMI), CGM metrics, user-reported outcome measures and frequency of clinical events are reported as secondary outcomes (Table 1). Glucose management indicators (GMI) will not be used in lieu of laboratory HbA_1c; GMI is captured as its own data point. Although analysis will be performed at intervals, as an audit of clinical care, follow-up frequency will be determined by the clinical teams responsible, based on clinical need. At follow-up the same clinical outcomes that were captured at baseline will be evaluated where available through patient reporting and review of clinical systems. Sensor glucometrics will be extracted from the relevant HCL system for the 14 days preceding any follow-up.

Data will be assessed for accuracy and completeness. Values thought to be erroneous will be ﬂagged for review at the centre submitting the data. Data will be cleaned and analysed using Stata SE 16. Analysis will utilise paired data from individuals with baseline and follow-up at the relevant time interval. The numbers with missing data at baseline and follow-up will be reported.

Stratiﬁed analysis by HbA_1c level, age, HCL system, type of insulin and ethnicity subgroups (for example) will be performed for each outcome. Where users have changed system or insulin the HCL system or insulin used at follow-up will be used for stratiﬁcation purposes. Following the initial analysis, further analyses will also include subgroup comparisons between those accessing HCL via the NHS England pilot and those using pre-existing criteria or any new criteria that may be announced in future.

Continuous and numerical variables including event rates, Gold Score and DDS2 will be assessed for normal distribution. Changes in normally distributed continuous covariates will be assessed using paired T-Tests. Wilcoxon Signed Rank tests will be used to assess changes in non-normally distributed data. Stratiﬁed analyses will be performed for these outcomes, utilising ANOVA for normally distributed variables or the Kruskal-Wallis test for non-normally distributed variables. Results for pairwise comparisons between subgroups will be Bonferroni-corrected. Comparisons between those switching systems and those remaining on a single system will be performed as a sensitivity analysis.

Adjustment of change in HbA_1c and weight from baseline for baseline characteristics and change in other covariates will be performed using a multiple linear regression model to correct for key covariates determined a priori as follows: baseline HbA_1c /weight, gender, age, duration of diabetes, deprivation level, HCL system and ethnicity.

The total number of clinical events (admissions, paramedic callouts, DKA and severe hypoglycaemia) and number of people experiencing these events, at baseline and follow-up (adjusted pro rata) will be compared using Chi-squared tests. Events per person/year rates will be calculated at baseline and follow-up to facilitate comparison. Mean Likert scores for user and caregiver opinions will be reported. The frequency of any reported adverse events will be reported.

Discussion

This ABCD clinical audit will be one of the largest independent audits of routine clinical outcomes to capture real-world data from multiple HCL systems. It builds on the broad expertise and experience of the ABCD audit programme, which has a record of providing novel insights from real-world clinical practice. The group included in the NHS England pilot are of particular interest because they are individuals with higher baseline HbA_1c levels, a group often not included in RCTs.

The strengths of this audit and proposed analyses lie in the well-tested design which will produce ﬁndings reﬂective of real-world practice. Local areas will also be able to access their own data to review their outcomes, improve standards and potentially to advocate for access to HCL technology in their area. This will be the ﬁrst independent audit to incorporate multiple different systems being used in the real world. Our initial analysis will be the ﬁrst to focus on a cohort with elevated HbA_1c levels at baseline, who are poorly represented by the current RCT and observational evidence. It will also capture a broad range of data, including assessment of clinical outcomes such as hospital admission rates and retinopathy data which are not currently reported in other real-world studies. Future analysis featuring a broader range of HCL users, accessing the technology through various criteria, will allow for greater generalisability but will also provide the opportunity to contrast the real-world outcomes in clinically different subgroups.

Despite these strengths, the clinical audit design of this work can introduce problems if there are incomplete or erroneous data. Regular review of the data will allow for troubleshooting of suspected erroneous values or missing key data to minimise this risk. Finally, whilst inclusion and analysis of outcomes in those on the NHS England HCL pilot will produce novel data, it may still fail to answer some questions. Chieﬂy, because the pilot is only accessible to those already on technology (FreeStyle Libre and an insulin pump) it will not provide an insight of the potential beneﬁts of taking someone with an elevated HbA_1c directly from multiple daily injections to HCL or from pump (without sensor) to HCL. However, inclusion of HCL users beyond the NHS pilot should overcome this limitation, by providing data from those transitioning from multiple daily injection therapy to HCL.

Conclusion

This audit programme has the potential to provide a large real-world dataset of HCL therapy in those living with T1D and will be key in informing the future roll-out of this technology. Whilst there are limitations to its design, it will provide a rich data set with a focus on those accessing technology via the NHS England pilot and beyond – groups from whom current data are limited. Ongoing adoption and input into the audit programme will allow future surveillance and reporting of HCL outcomes across the UK and will allow us to compare those accessing the technologies via multiple different criteria.

Conﬂict of interest TSJC has received speaker fees and/or support to attend conferences from NovoNordisk, Sanoﬁ and Abbott Diabetes Care. TPG has received personal fees from NovoNordisk, Sanoﬁ Aventis, Mundipharma Pharmaceuticals, Dexcom, Abbott Diabetes Care and Eli Lilly. EGW has received personal fees from Abbott Diabetes Care, Dexcom, Eli Lilly, Glooko, Insulet, Medtronic, Novo Nordisk and Sanoﬁ Aventis. PC has received personal fees from Abbott Diabetes Care, Dexcom, Eli Lilly, Insulet, Medtronic, Novo Nordisk, Sanoﬁ Aventis and Glooko. AL has received personal fees from Abbott Diabetes Care, Dexcom, Insulet, NovoNordisk, Sanoﬁ Aventis and research support from Abbot Diabetes Care and Novo Nordisk.

Funding This audit is funded by the Association of British Clinical Diabetologists.