Automated closed-loop insulin delivery for the management of type 1 diabetes during pregnancy: the AiDAPT RCT

Article history

The research reported in this issue of the journal was funded by the EME programme as award number 16/35/01. The contractual start date was in January 2018. The final report began editorial review in April 2023 and was accepted for publication in November 2023. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ manuscript and would like to thank the reviewers for their constructive comments on the final manuscript document. However, they do not accept liability for damages or losses arising from material published in this manuscript.

Copyright statement

Copyright © 2024 Lee et al. This work was produced by Lee et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Lee et al.

Overall trial design

Automated insulin delivery among pregnant women with type 1 diabetes (AiDAPT) was a multicentre, randomised, open-label, two-arm parallel-group trial comparing automated hybrid closed-loop and CGM alongside standard insulin delivery in pregnant women with type 1 diabetes.

Pregnant women with at least 1 year’s duration of type 1 diabetes who were ≤ 13 weeks and 6 days’ gestation with an early‑pregnancy HbA1c of 48 to ≤ 86 mmol/mol (6.5 to ≤ 10.0%) were approached by local clinical teams. Participants were recruited through nine outpatient antenatal diabetes clinics in NHS maternity clinics across England (Norwich, Ipswich, Cambridge, Leeds and two London sites), Scotland (Glasgow and Edinburgh) and Northern Ireland (Belfast). At enrolment all participants were asked to complete a run-in phase using the study CGM (Dexcom G6 system, Dexcom, Inc., San Diego, CA, USA) to collect baseline maternal glucose data and assess tolerance to wearing the devices before randomisation. Participants who were already using the Dexcom G6 CGM system before enrolment continued using it unmasked during the run-in period. Participants using other intermittently scanned (Freestyle Libre, Abbott Diabetes Care, Alameda, CA, USA) or other CGM systems were given a masked Dexcom G6 CGM. Participants using off-label closed-loop systems other than CamAPS FX were eligible provided they were willing to use the trial CGM (Dexcom G6) system.

Those for whom ≥ 96 hours of CGM data with ≥ 24 hours overnight (23.00–07.00) were collected were randomised on a 1 : 1 basis to continue with CGM alongside standard insulin delivery, which was either MDIs or insulin pump therapy, or to use the study hybrid closed-loop system. Training (inperson or virtual) was provided by local teams on the Dexcom G6 sensor insertion, CGM data interpretation, dietary advice and insulin dose adjustment (see Report Supplementary Material 1). Participants randomised to hybrid closed-loop used the same CGM (Dexcom G6) system with an insulin pump (Dana Diabecare RS, Advanced Therapeutics UK Ltd., Warwick, UK) and control algorithm (CamAPS FX), hosted on a mobile phone app. The closed-loop devices used in the trial are shown in Figure 1. The system was designed and implemented using appropriate cybersecurity approaches and has been used securely both in research and in clinical practice without any cybersecurity breaches or concerns. Participants were alerted by a ‘systems alarm’ if closed-loop stopped working, for example during loss of connection between the phone and insulin pump. For safety, this alarm was mandatory and could not be turned off. Data from the closed-loop system (CGM and insulin pump data) were stored on a cloud-based system (Glooko/Diasend, Glooko, Inc., Palo Alto, CA, USA). The linked-anonymised closed-loop system was shared with the trial research teams and with antenatal teams at local sites.

FIGURE 1.

Hybrid closed-loop system.

A training session (inperson or virtual) covering using the closed-loop devices, alarms and troubleshooting was provided by the study research educator or local care team within 2–4 weeks after randomisation.

Participants in both arms used the same Dexcom G6 CGM system with support for insulin dose adjustment from their usual antenatal clinical care team. For participants in the standard care group, the anonymised CGM data were recorded by the device manufacturer’s web-based diabetes management software (Dexcom CLARITY^®). The linked-anonymised CGM data were shared with the trial research teams and with antenatal teams at local sites. The CGM system has alarms to alert about actual or impending high or low glucose excursions which are user specified. For safety, the low glucose alarm (which is set at 3.3 mmol/l) was mandatory and could not be turned off. Participants could also choose whether to share their personal CGM data with significant others (partners, family members etc.).

Participants could continue to use the study devices during antenatal hospital admissions, including the delivery admission, and on the postnatal hospital ward according to local clinical guidelines.

Modifications to the study protocol

Two events, the COVID-19 pandemic and changes in NICE diabetes pregnancy clinical guidelines, led to modifications to the original study protocol.

Patient and public involvement

Patients and public representatives were involved in the design of the trial as co-applicants, and as part of the TSC group, throughout the conduct of the trial and involved in all key protocol amendment decisions before, during and following the COVID-19 pandemic, as well as contributing to the interpretation and dissemination of results.

Primary research questions

(Data from Lee et al. 36) Among pregnant women with type 1 diabetes:

1. What is the biomedical impact of using automated hybrid closed-loop insulin delivery throughout pregnancy?

   1. Does hybrid closed-loop use improve maternal glucose outcomes during the second and third trimester, compared to standard insulin delivery?

   2. Is hybrid closed-loop use safe in terms of rates of maternal hypoglycaemia, DKA and AEs?

   3. Is in-hospital use of automated hybrid closed-loop safe on general obstetric wards and delivery units?

2. What is the psychosocial impact of using automated hybrid closed-loop insulin delivery?

   1. What are women’s experiences of, and views about, using a hybrid closed-loop insulin system to manage their diabetes during pregnancy?

   2. How might hybrid closed-loop systems be improved for future use by pregnant women?

   3. What information, training and support do healthcare professionals need to support pregnant women to optimally use hybrid closed-loop systems?

Potential participants were identified by treating clinicians, provided with study information leaflets either in person or by post/e-mail and invited to join the study usually at least 1 week before the recruitment visit. Participants were eligible for recruitment if they fulfilled the following inclusion criteria.

Inclusion criteria

1. Between 18 and 45 years of age

2. Type 1 diabetes for at least 12 months’ duration

3. Viable pregnancy confirmed by ultrasound, up to 13 weeks and 6 days’ gestation

4. On intensive insulin therapy (three or more injections/day or insulin pump). This included sensor-augmented insulin pumps and hybrid closed-loop systems other than CamAPS FX

5. Willingness to use the study devices throughout the trial

6. HbA1c level ≥ 48 mmol/mol (≥ 6.5%) at booking (first antenatal contact) and ≤ 86 mmol/mol (≤ 10%) at point of randomisation. A CGM or Libre GMI ≥ 48 mmol/mol (≥ 6.5%) or ≤ 86 mmol/mol (≤ 10%) was used if laboratory HbA1c could not be obtained37

7. Provided written informed consent

8. Had access to an e-mail account

Exclusion criteria

1. Non-type 1 diabetes

2. Other physical or psychological disease which was likely to interfere with the normal conduct and interpretation of the study results, as judged by the site investigator

3. Current treatment with drugs known to interfere with glucose metabolism (e.g. high-dose corticosteroids)

4. Known or suspected insulin allergy

5. Advanced nephropathy (estimated glomerular filtration rate < 45), severe autonomic neuropathy, uncontrolled gastroparesis or severe proliferative retinopathy, as judged by the site investigator

6. Target glycaemia or very high HbA1c that is first antenatal HbA1c < 48 mmol/mol (< 6.5%) and HbA1c > 86 mmol/mol (> 10%). Those with HbA1c > 86 mmol/mol (> 10%) may participate if they achieve HbA1c ≤ 86 mmol/mol (≤ 10%) before randomisation

7. Total daily insulin dose > 1.5 units/kg suggesting severe insulin resistance

8. Severe visual or hearing impairment

9. Unable to speak and understand English

Recruitment visit

The following activities were performed before 14 weeks’ gestation:

• Checking for inclusion and exclusion criteria

• Written informed consent

• Past medical (diabetes and obstetric) history

• Body weight and height, calculation of body mass index (BMI)

• Baseline questionnaire pack provided for participants to complete at home (either paper or electronically via link)

• Dexcom G6 sensor insertion.

Written informed consent was obtained by trained staff at each site before any study-specific procedures. Baseline data included past medical, diabetes and obstetric history, current diabetes management and a brief physical examination. Participants were asked to complete the following validated questionnaires: EuroQol-5 Dimensions health-related quality-of-life questionnaire (EQ-5D),38 Diabetes Distress Scale (DDS),39 HFS II (worry scale only),40,41 Pittsburgh Sleep Quality Index (PSQI),42 and the Insulin Delivery Systems: Perceptions, Ideas, Reflections and Expectations (INSPIRE) measure. 43

Participants used a Dexcom G6 study CGM device during the run-in phase to provide a baseline assessment of maternal glycaemia (at least 96 hours of glucose values, including 24 hours overnight) and to ensure that the device was tolerated. Baseline glucose values were masked except for those already using the Dexcom G6 trial continuous glucose monitor.

For participants not already using Dexcom G6, a Dexcom G6 subcutaneous glucose sensor was inserted by the clinical research team and the participant was instructed to wear it at home for up to 10 days with a receiver device. They were asked to return the receiver for uploading of their anonymised baseline CGM data within 14 days.

For participants who were already using Dexcom G6, they were asked to insert a new Dexcom G6 glucose sensor and either given a receiver device as above or switched from their personal Dexcom account to an anonymised study Dexcom account on their smartphone.

Randomisation visit

After the run-in phase, eligible participants underwent randomisation to either hybrid closed-loop or standard insulin delivery with CGM.

The following activities were performed before 16 weeks’ gestation:

• CGM sensor upload from receiver/CGM data review

• Baseline bloods (where possible)

• Collection/confirmation of the completed baseline questionnaires

• Confirm HbA1c or GMI level ≤ 86 mmol/mol (10%)

• Record average total daily dose of insulin during the previous 3 days

• Randomisation via study website

• Participant training

The CGM sensor data were downloaded from the receiver by the research team and/or reviewed via the study account to provide a baseline glucose control assessment. At least 96 hours of CGM glucose values with 24 hours of glucose values during the hours of 23.00 to 07.00 were required. If there were any technical difficulties and/or inadequate CGM data, a second CGM sensor was provided (if possible, within the required time frames for visits). The CGM readings recorded during this period were also used to optimise insulin therapy in both groups.

If laboratory measurements of HbA1c levels were unavailable (e.g. due to COVID-19 regulations), estimates from GMI in Libre/CGM were acceptable.

Treatments were allocated in a 1 : 1 ratio via a web-based system hosted by the Jaeb Center for Health Research, which used a computer-generated randomisation list with permuted block sizes of two to four, and stratification by clinical site.

Participant technical support

All participants had access to support from their local site teams and Dexcom technical support in case of technical problems with their CGM devices and connectivity. Those randomised to closed-loop insulin delivery were also signposted to Advanced Therapeutics in case of insulin pump-related problems and had access to a telephone helpline to contact the research study team for any concerns about their closed-loop function and device connectivity.

Treatment discontinuations

In consenting to the trial, participants consented to trial treatments, trial follow-up and data collection. However, an individual participant could decide to stop treatment early or could be stopped early for any of the following reasons:

• Unacceptable adverse device effect or AE

• Intercurrent illness that prevented further treatment

• Any change in the participant’s condition that in the clinician’s opinion justified the discontinuation of treatment

• Withdrawal of consent for treatment by the participant

• Significant Clinical Investigation Plan violation or non-compliance

• Allergic reaction to insulin

• Technical problems with the closed-loop system, which could not be resolved

• Any other significant medical event or start of medications that significantly affected glucose metabolism (with the exception of prophylactic steroids for fetal lung maturation).

As participation in the trial was entirely voluntary, the participant could choose to discontinue trial treatment at any time without penalty or loss of benefits to which they would otherwise be entitled. Although not obliged to give a reason for discontinuing their trial treatment, every reasonable effort was made to establish the reason(s) while remaining fully respectful of the participant’s rights. Participants who discontinued treatment, for any reason, remained in the trial for the purpose of follow-up, and data collection and analysis, if they were willing.

Follow-up visits and data collection

Ongoing study visits were scheduled to coincide with routine clinic visits, which occurred at least 4-weekly from 12 to 36 weeks’ gestation. Maternal weight and blood pressure, insulin dose and type, details of device issues and AEs were recorded at each study visit. Participants were asked to repeat the baseline questionnaires at 34–36 weeks’ gestation with an additional INSPIRE questionnaire for intervention-arm participants. 43 CGM and insulin data were collected via the manufacturer’s cloud software. Obstetric input and ultrasound scans were performed at approximately 20, 28, 32 and 36 weeks’ gestation as per NICE guidelines for pregnant women with type 1 diabetes. Any inpatient hospital admissions were recorded. At delivery, data regarding obstetric and neonatal outcomes were collected.

Maternal and neonatal outcomes

For pregnant women, prespecified health outcomes were: gestational weight gain (change in maternal weight between the initial antenatal visit and the final trial visit before delivery, typically 34–36 weeks’ gestation), gestational hypertension, pre-eclampsia, mode of delivery and maternal length of hospital stay. Gestational hypertension was defined as systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg on at least two occasions 4 hours apart, developing after 20 weeks’ gestation in previously normotensive women. Pre-eclampsia was hypertension accompanied by proteinuria ≥ 300 mg in 24 hours, or two readings of at least ++ on dipstick analysis of urine or documentation of pre-eclampsia in the delivery or antenatal hospital records.

Neonatal health outcomes included common diabetes-related complications of preterm delivery, neonatal hypoglycaemia, large for gestational age and neonatal care unit admission, as well as other clinically important adverse pregnancy outcomes which occur infrequently: shoulder dystocia, birth injuries and neonatal death. We defined preterm delivery as birth before 37 weeks’ gestation and neonatal hypoglycaemia as a capillary glucose measurement of < 2.6 mmol/l on one or more occasions within the first 48 hours of life, starting at least 30 minutes after birth and necessitating treatment either with 40% glucose gel administered to the buccal mucosa and/or with i.v. dextrose. Neonatal care unit admission included any admission with a duration of at least 24 hours that required separation of mother and baby. Some NHS maternity units provide transitional care units where the parents are the primary caregivers and only minimal staff support is required. We therefore categorised three different levels of neonatal care unit admissions. Level 1 neonatal care (also called special care baby unit) was for babies who needed continuous monitoring of their breathing or heart rate, additional oxygen tube feeding, phototherapy recovery (to treat neonatal jaundice) and convalescence from higher-level neonatal care units. Neonatal hyperbilirubinaemia was defined as significant jaundice based on bilirubin levels requiring treatment with either phototherapy > 6 continuous hours, or an exchange transfusion, or receiving i.v. gamma globulin or requiring re-admission into hospital during the first 7 days of life due to hyperbilirubinaemia. Level 2 neonatal care was for babies needing short-term intensive care and typically includes those with apnoeic attacks who require respiratory support, including continuous positive airway pressure. Some babies receiving parenteral nutrition or i.v. dextrose may also need level 2 neonatal care. Level 3 or NICU is used for the most unwell babies, typically those delivered preterm and/or needing respiratory support, or other high-level care. Respiratory distress was defined as respiratory difficulties requiring any positive-pressure ventilation ≥ 24 hours, beyond resuscitation period (10 minutes) and/or given surfactant within 72 hours after birth. The duration of neonatal care and total length of neonatal hospital stay were also recorded.

Because of differences in gestational age at delivery, in addition to neonatal birthweight, and macrosomia (defined as birthweight ≥ 4 kg), we used gestation-related optimal weight birthweight centiles that adjust for both neonatal (sex and gestational age) and maternal factors (height, weight, parity and ethnicity). These were used to calculate the birthweight percentile and the proportion of infants that were large or small for gestational age (defined as birthweight percentile > 90th or < 10th percentile, respectively).

We used a composite measure of pregnancy loss including any of miscarriage, stillbirth or neonatal death, as well as individual miscarriage, stillbirth or neonatal death outcomes. We defined stillbirth as a fetal loss occurring after 24 weeks’ gestation, and neonatal death as death of a live-born infant up to 28 days after delivery. To capture other clinically important adverse pregnancy outcomes, we recorded all episodes of shoulder dystocia and all serious birth injuries. Shoulder dystocia was defined as a vaginal cephalic delivery that required additional obstetric manoeuvres to deliver the fetus after the head had delivered and gentle traction had failed. Birth injuries included any of the following: spinal cord injury, basal skull fracture or depressed skull fracture, clavicular fracture, long bone fracture (humerus, radius, ulna, femur, tibia or fibula), subdural or intracerebral haemorrhage of any kind (confirmed by cranial ultrasound, computerised tomography scan or magnetic resonance imaging), peripheral nerve injury/brachial plexus.

Safety outcomes

Participants were reviewed for use of study devices and AEs including a skin assessment at each visit. All AEs and device deficiencies were recorded on a web-based database, with additional detail on SAEs, SADEs, DKA events, severe hypoglycaemic events and inpatient admissions. We collected data on episodes of severe hypoglycaemia (SH), ketosis and DKA from the time of recruitment until the date of discharge from hospital after giving birth. SH was defined as an event requiring assistance of another person actively to administer carbohydrate, glucagon or other resuscitative actions. SH events were further categorised as treated at home with rescue carbohydrates and/or glucagon, requiring ambulance or paramedic call-out and requiring hospital admission. Measures of acidosis (pH and/or bicarbonate), peak glucose levels, hospital admission status and glycaemic management were recorded where available. Ketosis was defined as ketones > 0.5 mmol/l. Episodes of hyperglycaemia with ketosis were categorised as mild to moderate if capillary and/or plasma ketone measures were > 0.5 mmol/l and they were self-treated and resolved without hospital admission. Episodes of hyperglycaemia with ketosis were categorised as severe if capillary and/or plasma ketone measures were > 1.0 mmol/l and hospital admission and treatment with variable rate i.v. insulin infusion were required. Episodes of DKA were identified using Joint British Diabetes Society thresholds (presence of diabetes mellitus of any kind; capillary ketones > 3 mmol/l and acidosis as per blood gas bicarbonate < 15 mmol/l or pH < 7.3), or if they were managed with fixed-rate i.v. insulin infusion. 44

Serious adverse events and SADEs were notified to Norwich Clinical Trials Unit and reported onward to the MHRA, device manufacturer and research ethics committee as required. Dexcom G6 and CamAPS FX app investigational devices were CE marked during the trial, and thereafter, MHRA was informed but did not require expedited notifications.

Bloods

Blood collection for HbA1c levels was performed where possible, at randomisation, 24–26 and 34–36 weeks at each site, using an International Federation of Clinical Chemistry Laboratory Medicine-aligned methodology, with optional biorepository samples for future metabolic studies in those who provided specific consent.

Qualitative interviews

Twenty-three participants randomised to closed-loop were recruited from across the trial sites and purposively sampled to capture diversity in terms of age, education, socioeconomic status, previous pregnancies, diabetes duration and baseline HbA1c. Baseline interviews were conducted post randomisation to enable pre-pregnancy diabetes management and initial women’s expectations of closed-loop to be explored. The same participants were reinterviewed at 34–36 weeks’ gestation to explore whether and how using hybrid closed-loop affected their diabetes management, pregnancy experiences, work and family lives.

Nineteen trial staff were recruited from across trial sites and sampled to capture diversity in clinical and trial experience. Interviews were conducted near the end of the trial with an additional online workshop to explore staff’s experiences of delivering the trial and supporting pregnant women using closed-loop insulin delivery, and their views about the training and resourcing health professionals would need to support women using closed-loop systems in routine clinical care.

The trial flow chart and visit schedules are shown in Figure 2.

FIGURE 2.

Flow of participants through the trial. ^aVirtual device training procedures and visits were implemented following the COVID-19 lockdown restrictions.

Trial end points

The primary efficacy end point was between-group difference in change in the percentage of time spent in the pregnancy target glucose range of 3.5–7.8 mmol/l from 16 weeks’ gestation until delivery. We defined the null hypothesis as no difference in time spent in the target glucose range (3.5–7.8 mmol/l) between those pregnant women who used hybrid closed-loop insulin delivery and those who used CGM alongside a standard insulin delivery method (insulin pump or MDIs) during the second and third trimester. The alternative hypothesis was a non-zero difference (two-sided) in time spent in the target glucose range (3.5–7.8 mmol/l) between women who used hybrid closed-loop insulin delivery and those who used CGM alongside a standard insulin delivery method (insulin pump or MDIs) during the second and third trimester.

Key secondary end points were percentage of glucose levels above target range, defined as > 7.8 mmol/l reflecting antenatal hyperglycaemia, and percentage of overnight time spent in the target glucose range, reflecting AID without the impact of insulin boluses. A prespecified subset of sensor glucose outcomes (mean glucose, percentage time spent in, above and below relevant thresholds, glycaemic variability and hypoglycaemic events) were calculated for overnight (23.00–07.00) and for each trimester. Additional secondary outcomes included glycated haemoglobin (HbA1c), insulin doses and attainment of sensor glucose targets. The prespecified subset of outcomes were all tested for superiority. Secondary outcomes were all listed in a prespecified statistical analysis plan and are summarised below.

Summary of trial outcomes

1. The percentage of time spent with sensor glucose levels above and below target range (> 7.8 and < 3.5 mmol/l), mean sensor glucose and glucose variability measures; glucose standard deviation (SD) and glucose coefficient of variation (CV).

2. The frequency and severity of hypoglycaemia episodes < 3.5 mmol/l (mild) and < 3.0 mmol/l (moderate) for more than 15 minutes’ duration.

3. The international CGM TIR consensus targets: CGM glucose levels 3.5–7.8 mmol/l > 70% (16 hours 48 minutes), > 7.8 mmol/l < 25% (6 hours), < 3.5 mmol/l < 4% (1 hour) and < 3.0 mmol/l < 1% (15 minutes).

4. The low blood glucose index (LBGI) to quantify the risk of hypoglycaemia.

5. Change in maternal HbA1c based on blood samples collected at baseline, 24–26 weeks and 34–36 weeks’ gestation.

6. CGM glucose levels during the first (< 12 weeks 6 days’ gestation), second (13–27 weeks 6 days’ gestation) and third trimesters (28 weeks until delivery).

7. CGM glucose levels during the 24 hours (midnight to midnight) and overnight (23.00–07.00).

Detailed trial outcomes

Continuous glucose monitoring and closed-loop system use

The amount of CGM use in the intervention group was calculated over the period starting from the day after hybrid closed-loop treatment started until the delivery date or date of pregnancy loss (miscarriage or pregnancy termination date). Participants who dropped out were counted as zero use from that point forward until the delivery date (if known) or the estimated delivery date (based on the ultrasound scan). Participants with pregnancy loss or preterm births were included up until the date of pregnancy loss/preterm birth when calculating percentage of CGM use.

The percentage of closed-loop system use in the intervention group was calculated in a similar manner. Box plots were created for percentage time closed-loop system use and percentage time CGM use overall, by day and night and by 4-weekly antenatal period (12–16, 16–20, 20–24, 24–28, 28–32, 32–36 and > 36 weeks’ gestation) in the intervention group. For the standard care control group, the amount of CGM data in the post-randomisation period throughout pregnancy was similarly summarised. Scatterplots were created for closed-loop system use versus selected CGM metrics and HbA1c at 34 weeks in the intervention group.

Statistical analyses

We calculated that we would need to enrol 98 participants to provide the trial with 90% power to detect a difference, assuming a true population difference of 10% absolute difference in the primary outcome (percentage time spent in the pregnancy-specific target glucose range) from 16 weeks’ gestation until delivery, based on a SD of 15% and a two-sided type 1 error rate of 5%. The sample size was increased to 124 to allow for dropouts due to pregnancy losses and withdrawals.

Statistical analyses were performed on an intention-to-treat basis including all participants with at least 96 hours of sensor glucose data between 16 weeks’ gestation and delivery. For the primary analysis, the groups were compared using a linear mixed-effects regression model adjusting for baseline TIR, insulin delivery and clinical site as a random effect. Missing primary end-point data were handled using multiple imputation with pattern mixture models (Rubins and direct likelihood methods) with all randomised participants included. For secondary outcomes, analyses were similar to the primary analysis, without imputation. False discovery rate (FDR)–adjusted p-values were calculated for selected secondary outcomes (overall, overnight, and by-trimester sensor glucose metrics, HbA1c, insulin doses, subgroup analyses, questionnaires) using the two-stage adaptive Benjamini–Hochberg method. For attainment of sensor glucose targets, a mixed-effects logistic regression model was fitted adjusting for baseline TIR, insulin delivery and clinical site as a random effect. All p-values are two-tailed. For exploratory analyses between closed-loop system use and maternal glucose outcomes (CGM TIR, HbA1c at 34 weeks), Spearman correlation was used. Analyses were performed using SAS^® 9.4 (SAS Institute Inc., Cary, NC, USA; SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ^® indicates USA registration).

Participants

Between September 2019 and May 2022, 334 participants were assessed for eligibility, with 208 excluded. The reasons for their exclusions included not meeting the eligibility criteria [n = 135: HbA1c out of range (n = 60); unwilling to use study devices or switch from current treatment methods (n = 32); outside of gestational age window (> 13 weeks 6 days) (n = 24); other reasons (n = 19)], declining to participate (n = 47), site or participant COVID-19-related reasons (n = 12), early‑pregnancy losses (n = 7) and other non-specified reasons (n = 7).

Two participants completed the run-in phase but subsequently declined to be randomised, one stating that she did not want to be randomised to hybrid closed-loop therapy and another who did not want to be randomised to the trial CGM system. The remaining 124 participants from nine NHS trial sites were randomised, 61 to hybrid closed-loop and 63 to CGM alongside standard care insulin delivery (Table 1 and Figure 3).

FIGURE 3.

CONSORT flow diagram. Reasons for not meeting trial eligibility criteria (N = 135) were: HbA1c out of range (n = 60), unwilling to use study devices or switch from current treatment methods (n = 32), outside of gestational age window (> 13 weeks 6 days; (n = 24), other reasons (n = 19).

Recruitment across trial sites

Leeds Teaching Hospitals NHS Foundation Trust had the smallest number of participants in the intervention group (n = 1), followed by Belfast Health and Social Care Trust (n = 2) which was the last trial site to join. The highest numbers of trial participants were recruited from the lead sites, Norfolk and Norwich and Cambridge University Hospitals NHS Foundation Trusts, which had 19 and 11 participants, respectively, in the closed-loop group.

One participant from Edinburgh was randomised to the closed-loop intervention group on the first day of the COVID-19 lockdown (17 March 2020), before virtual training procedures were implemented, and was therefore switched by the research team to CGM and standard insulin therapy. One participant from Cambridge who was randomised to the standard care control group procured the CamAPS FX closed-loop system outside the trial procedures (Figure 3).

Participants were aged 20–45 years, with BMI ranging from 18 to 49 kg/m², and most had long duration of type 1 diabetes (17 ± 8 years) and self-identified as having white ethnic/racial heritage (Table 2). Their median gestational age at recruitment was 10 weeks, with randomisation occurring 1 week later at approximately 11 weeks’ gestation. In terms of education, 40% of participants had a university undergraduate degree or equivalent, while a smaller proportion (16%) had a postgraduate degree or equivalent.

Participants’ mean HbA1c measurement, which was taken as soon as possible after pregnancy confirmation, was 61 mmol/mol (7.7%). Almost all (98%) trial participants were using intermittently scanned (74% Freestyle Libre) or real-time CGM (26%) at enrolment, meaning that most were performing fewer than three daily self-monitoring of capillary blood glucose (SMBG) checks (Table 3). Approximately half were using MDIs and half were using insulin pump therapy, with three participants using alternative commercially available AID systems [one DIY loop Android APS via Accu-Chek Insight (Roche Diabetes Care, Inc., Basel, Switzerland), one Tandem Control IQ, one Medtronic 780G] during early pregnancy.

Approximately half of the trial participants were taking pre-conception folic acid, suggesting that they were actively preparing for pregnancy. The group had relatively high rates of diabetes complications, with 70 participants (56%) reporting previous retinopathy, neuropathy or nephropathy, and a substantial proportion who reported smoking cigarettes (19%) and drinking alcohol (58%) before pregnancy. Participants in the standard care group reported more DKA events in the 12 months before pregnancy.

Approximately one-third of trial participants reported previous pregnancy losses, either miscarriage or termination of pregnancy (Table 4). Participants in the closed-loop group had more previous births. Six trial participants (5%) had chronic hypertension diagnosed prior to pregnancy.

Blood pressure measurements at enrolment were limited during the pandemic, with missing data for five participants in the closed-loop group and nine participants in standard care. Any minor imbalances (more previous DKA events in the standard care group, higher parity in the closed-loop group) were within the expected bounds for random allocation.

Two participants in each group did not contribute to the primary efficacy end point. In the closed-loop group, this included one participant with a first-trimester miscarriage who had < 96 hours of CGM data and one withdrawal of a previous closed-loop (Tandem Control IQ) user. Similarly, there was one first-trimester miscarriage with < 96 hours of CGM data in the standard care group and one participant who declined to use the trial CGM system. Overall, seven participants in each group discontinued their allocated treatment, the timing and reasons for which are listed in Table 5.

Reasons for having < 96 hours’ CGM data in the intervention group were one miscarriage < 12 weeks and one withdrawal of a previous closed-loop (Tandem Control IQ) user 17 days post training. Reasons in the standard care group were one miscarriage < 12 weeks and one withdrawal of a previous Freestyle Libre user before CGM training.

Reasons for not completing the 34–36 weeks visit (if not delivered by then) in the closed-loop group were one miscarriage < 12 weeks and one withdrawal of previous Tandem Control IQ user. Reasons in the standard care group were one miscarriage < 12 weeks, one withdrawal of a previous Freestyle Libre user and two pregnancy terminations (one for congenital anomaly). Three participants in the standard care group completed the 34–36 weeks’ gestation visit but discontinued CGM in late pregnancy (resumed use of Freestyle Libre) due to device connectivity issues, bringing the total number of discontinuations to seven.

Reasons for closed-loop being active < 60% of the time in the intervention group were: one miscarriage < 12 weeks’ gestation; one intervention group participant who was switched to standard care by the research team on day 1 post randomisation (17 March 2020) due to COVID-19 lockdown restrictions; one withdrawal with no closed-loop use from 16 weeks’ gestation until delivery due to deteriorating medical and mental health comorbidities (20 hyperemesis and severe ketosis events); and four withdrawals at days 15, 17, 17 and 21 post device training from participants who stated that the CamAPS closed-loop system was not sufficiently aggressive/responsive [these included the previous closed-loop (Tandem Control IQ) user with entry HbA1c 6.0%].

Despite the impact of the COVID-19 pandemic on trial participants and on NHS healthcare teams, the proportion of completed study visits was high: approximately 95% from 16 weeks’ gestation (Figure 4).

FIGURE 4.

Completion of trial visits. ^aParticipants who were randomised before 12 weeks. Study visits were 4-weekly, so those randomised at 9–11 weeks did not require an additional 12-week visit. ^bParticipants who had not delivered prior to the 34–36 weeks visit. ^cFive participants (two intervention, three control) did not have 96 hours of sensor data between 16 weeks and delivery.

Participants in the standard care group had more additional clinic visits (1.5 vs. 1.1 per participant) and more unscheduled contacts (9.6 vs. 6.1 per participant), mostly for pregnancy- and diabetes-related reasons (Tables 6 and 7).

The reasons for additional unscheduled visits and contacts are listed in Table 7.

The frequency of trial CGM (Dexcom G6) sensor use was consistently high (median 97% across both treatment groups) and stable across pregnancy. Figure 5 shows side-by-side box plots of the sensor use for each treatment group, by 4-week antenatal period following device training.

FIGURE 5.

Frequency of CGM use by treatment group throughout pregnancy. Note: Black bars denote medians and black dots denote means. AID refers to hybrid closed-loop.

The situation was similar for the frequency of closed-loop system use in the intervention group, which was high (median 96%) and remained > 95% throughout pregnancy. Figure 6 shows box plots of the CamAPS FX closed-loop system use in the intervention group, by 4-week antenatal period following device training.

FIGURE 6.

Frequency of closed-loop use in the intervention group throughout pregnancy. Note: Black bars denote medians and black dots denote means.

Primary efficacy end point

The mean (± SD) percentage of time that maternal glucose levels were within the pregnancy target range from 16 weeks’ gestation until delivery increased from 47.8 ± 16.4% to 68.2 ± 10.5% in the closed-loop group and from 44.5 ± 14.4% to 55.6 ± 12.5% in the control group (mean‑adjusted difference 10.5 percentage points, 95% CI 7.0 to 14.0 percentage points; p < 0.001) (Table 8 and Figures 7–9).

FIGURE 7.

Cumulative distribution of percentage time spent in the target glucose range.

FIGURE 8.

Percentage time spent in the target glucose range throughout pregnancy. Note: Black bars denote medians and black dots denote means.

FIGURE 9.

Percentage time spent in the pregnancy target glucose range by time of day.

Figure 7 shows the cumulative distribution of the percentage of time that the glucose level was within the pregnancy-specific target glucose range of 3.5–7.8 mmol/l, as measured by CGM, for the hybrid closed-loop AID and standard care treatment group from 16 weeks’ gestation to delivery.

Figure 8 shows side-by-side box plots of the percentage of time that the glucose level was within the pregnancy-specific target glucose range of 3.5–7.8 mmol/l, as measured by CGM, for the hybrid closed-loop AID and standard care treatment group, over 4-week time periods from device training until delivery.

Figure 9 shows an envelope plot of the time spent in the same pregnancy-specific target glucose range (3.5–7.8 mmol/l), as measured by CGM, for each treatment group, according to the time of day, from 16 weeks’ gestation to delivery. The between-treatment-group difference is apparent across the 24-hour day but is most marked overnight, between 04.00 and 09.00.

Adjustment for potential confounding variables including previous pregnancies and DKA episodes prior to enrolment did not change the treatment difference (Table 9).

The large treatment difference was consistent between the intention-to-treat and the prespecified per-protocol analyses and using multiple imputation methods (see Tables 9 and 10). The treatment difference of 12 percentage points in those who used the closed-loop system for at least 60% of the time equates to an additional 3 hours per day spent in the pregnancy target glucose range from 16 weeks’ gestation until delivery.

We did not recruit any participants in second or subsequent pregnancies, and, therefore, a sensitivity analysis for first-pregnancy patients was not performed.

Secondary maternal glucose outcomes

Participants randomised to the closed-loop group also spent less time with glucose levels above target range, defined as > 7.8 mmol/l (mean difference −10.2%, 95% CI −13.8% to −6.6%; p < 0.001) (Table 11). This was accompanied by decreased hyperglycaemia across milder (> 6.7 mmol/l) and more pronounced (> 10.0 mmol/l) categories, as well as lower mean glucose and lower HbA1c (mean difference −0.31%, 95% CI −0.50% to −0.12%; p < 0.002).

The effects of the intervention during the overnight period (23.00–07.00) closely followed the 24-hour results (12.3% higher TIR; 95% CI 8.3% to 16.2%, p < 0.001). This was accompanied by a lower mean glucose concentration, less nocturnal hypoglycaemia, less glycaemic variability and fewer nocturnal hypoglycaemic events in the closed-loop group.

These changes are notable since participants in both groups spent approximately 70% of their time in the near-optimal glucose range of 3.5–10.0 mmol/l at baseline. Furthermore, in those who started closed-loop and CGM alongside standard care insulin therapy during the first trimester, a 5% higher time in the pregnancy-specific target glucose range was observed in the closed-loop group by 12 weeks 6 days’ gestation (Figure 8 and Table 12). The between-group treatment difference favouring closed-loop therapy increased to 12 percentage points (equivalent to spending an additional 3 hours per day in the pregnancy-specific target glucose range) during the second trimester and persisted thereafter. Changes in other CGM glucose metrics also favouring closed-loop therapy – mean glucose, time spent hyperglycaemic, glucose SD – remained clinically and statistically significant throughout the second and third trimesters.

The maternal glucose outcomes based on laboratory HbA1c measures at 34–36 weeks’ gestation also favoured the closed-loop treatment group, with mean HbA1c falling to a nadir of 6.0% at 24–26 weeks’ gestation and remaining at 6.0% during the third trimester. HbA1c rose to 6.4% during the third trimester in the standard care group (Table 13).

Attainment of the sensor glucose target of > 70% time (16 hours 48 minutes) within the pregnancy-specific target range of 3.5–7.8 mmol/l was achieved by 28 (47%) closed-loop and 7 (11%) standard care participants. Attainment of the sensor glucose target of < 25% time (6 hours) spent hyperglycaemic (> 7.8 mmol/l) was also achieved by more closed-loop participants; 22 (37%) closed-loop compared to 7 (11%) standard care participants (Table 14). Most participants in both treatment groups spent < 4% of the time (1 hour/day) with glucose levels below 3.5 mmol/l and < 1% of the time (15 minutes/day) with sensor glucose levels below 3.0 mmol/l.

Maternal glucose improvements were achieved without additional total daily insulin dose (Table 15).

Baseline questionnaires were completed by 116 participants (57 closed-loop, 59 standard care) with notably lower completion of the PSQI both at baseline and at follow-up. Overall, approximately two-thirds of trial participants completed the follow-up questionnaires, with no significant differences in any of the patient-reported outcomes between the two treatment groups (Table 16).

Maternal and neonatal outcomes

In addition to the striking improvements in maternal glucose outcomes, there was 3.7 kg less gestational weight gain in the closed-loop group (Table 17). We also observed less new-onset hypertension and more repeat caesarean sections in the closed-loop group, likely related to their higher number of previous pregnancies. There was one first-trimester miscarriage in the closed-loop group and one first-trimester miscarriage and two pregnancy terminations (one of which was for major congenital anomaly) in the standard care group.

In terms of serious birth injuries (Table 18), there was one shoulder dystocia (where the baby’s shoulders get stuck behind the mother’s pubic bone, requiring additional manoeuvres to release the shoulders during delivery) in the closed-loop group. This baby (male, weight 4.5 kg, gestational age 38 weeks 1 day) also had neonatal hypoglycaemia treated with i.v. dextrose and had a 6-day stay in a level 3 NICU. There were four serious birth injuries in the standard care group, one of which resulted in a neonatal death. This baby (female, weight 2.03 kg, gestational age 31 weeks 5 days) had severe hypoxic ischaemic encephalopathy (HIE) and died approximately 12 hours after birth by emergency repeat caesarean section following spontaneous onset of preterm labour. The participant declined a vaginal birth. There were three further HIE events in the standard care group, the most severe of which occurred in a baby (male, weight 4.17 kg, gestational age 37 weeks 1 day) with shoulder dystocia, fractured humerus, left Erb’s palsy and neonatal hypoglycaemia with an unrecordable blood glucose level at birth. This baby received immediate resuscitation, with therapeutic cooling, and spent 35 days in a level 3 NICU. A third baby (male, weight 3.31 kg, gestational age 37 weeks 1 day) in the standard care group had suspected HIE treated with positive-pressure ventilation as well as shoulder dystocia, Erb’s palsy, neonatal hypoglycaemia, neonatal jaundice and an 8-day NICU stay. The fourth baby (female, weight 3.68 kg, gestational age 37 weeks 2 days) had mild HIE treated with therapeutic cooling and neonatal hypoglycaemia with a 4-day level 3 NICU stay.

Babies of mothers in the closed-loop group were delivered 4.5 days earlier, without significant differences in common neonatal complications attributed to prematurity, including respiratory distress, neonatal hypoglycaemia, neonatal jaundice or overall rate of NICU admissions. However, babies of mothers in the closed-loop group did spend an additional 1 day longer in hospital, most likely due to their earlier gestation.

Safety outcomes

There were no unanticipated safety concerns associated with starting or using closed-loop during pregnancy. There were six SH events in the closed-loop group and five in standard care (Table 19). There was one clinical DKA event in each group. These were defined using Joint British Diabetes Society thresholds (capillary ketones > 3 mmol/l and acidosis as per blood gas bicarbonate < 15 mmol/l or pH < 7.3) or if they were managed with fixed-rate i.v. insulin infusion. 44 There were many more non-acidotic severe ketosis events (plasma ketones > 1.0 mmol/l) in the closed-loop group: 25 severe ketosis compared to 2 in the standard care group. The majority of these occurred in one participant with severe hyperemesis who experienced 20 out of the 25 non-acidotic ketosis events. Due to her deteriorating medical and mental health comorbidities she did not use closed-loop at any time between 16 weeks’ gestation and delivery but contributed to more ketosis and SAEs in the closed-loop group.

Seven participants each experienced one adverse device effect reported by local site teams as possibly, potentially or definitely related to CamAPS FX (Table 20). Five of these were related to possible device deficiencies, although unless associated with a SAE devicerelatedness was not reviewed by the trial team. One device deficiency with a SAE occurred; this severe hypoglycaemic event occurred after an early-pregnancy miscarriage during an inpatient hospital admission which also had additional causal factors including known epilepsy. The remaining events included one hyperglycaemia which contributed to a participant stopping closed-loop at day 17 post randomisation and one moderate ketosis event following overnight loss of Bluetooth connectivity the day before admission for preterm birth. Other events relating to sensor and/or infusion set failures were non-serious. The rate of adverse device events for the closed-loop system was 24.3 per 100 person-years.

Exploratory analyses

We also explored the correlations between the frequency of closed-loop system use and a selected range of maternal glucose outcomes, namely percentage time spent in the pregnancy-specific target glucose range of 3.5–7.8 mmol/l (Figure 10), HbA1c at 34–36 weeks’ gestation (Figure 11), time spent above the pregnancy target glucose range of > 7.8 mmol/l (Figure 12), time spent below the pregnancy target glucose range of < 3.5 mmol/l (Figure 13) and mean CGM glucose (Figure 14). While system use was generally very high, there were clear correlations between increased use and improved maternal glucose outcomes, with higher percentage time spent in the pregnancy target glucose range, lower HbA1c at 34–36 weeks’ gestation, less time spent above target range and lower mean glucose concentration. The time spent below target range was very low, with no meaningful correlation with closed-loop use.

FIGURE 10.

Scatterplot for percentage time spent in the pregnancy target glucose range and closed-loop system use.

FIGURE 11.

Scatterplot for maternal HbA1c at 34–36 weeks’ gestation and closed-loop system use.

FIGURE 12.

Scatterplot for percentage time spent above the pregnancy target glucose range and closed-loop system use.

FIGURE 13.

Scatterplot for percentage time spent below the pregnancy target glucose range and closed-loop system use.

FIGURE 14.

Scatterplot for mean sensor glucose concentration and closed-loop system use.

Study 1: Pregnant women’s experiences of closed-loop system use

Study 2: Experiences and views of trial staff providing care and support to participants in the closed-loop group

Final conclusions

Closed-loop was effective in type 1 diabetes pregnancy, safely accommodating the marked gestational changes in insulin doses across a range of maternal bodyweights and glycaemic categories. It gave additional clinical advantage above and beyond that which can be achieved by CGM and standard insulin therapy, supporting NICE guideline recommendations that hybrid closed-loop therapy should be offered to all pregnant women with type 1 diabetes.

Equality, diversity and inclusion

Impact, outputs and dissemination

An extensive multistranded dissemination strategy is under-way to ensure that the AiDAPT trial outputs reach widespread scientific and lay audiences as well as relevant stakeholders at NICE and NHS England. The primary manuscript has been published in the New England Journal of Medicine. 73 The data were submitted as ‘academic in confidence’ to the NICE Diagnostics Assessment team, for the assessment of hybrid closed-loop systems for managing blood glucose levels in type 1 diabetes www.nice.org.uk/guidance/indevelopment/gid-ta10845.

Acknowledgements

We are grateful to all the trial participants and their partners who have enthusiastically supported our research. https://www.heraldscotland.com/news/19841711.scots-doctor-fiona-denison-took-life-covid-devastated-health/ We would also like to acknowledge our Patient Public Involvement contributors, Mrs Sarah Cains and Ms Goher Ayman, for their invaluable input via the TSC throughout the performance of the trial. The trial was co-coordinated by Norwich Clinical Trial Unit (Norwich, UK) and the Jaeb Center for Health Research (Florida, USA), and we thank Jessica Rusnak (administrator, Jaeb Center) for her contributions to coordinating the teams’ collaborations. For assistance with the complex legal, financial and contractual issues, we sincerely thank Mercedes Mills (Research and Innovation Services, University of East Anglia, Norwich, UK). Finally, we thank Professor Jason Gardosi and colleagues at the UK Perinatal Institute (Birmingham, UK) for granting access to the latest version (gestation-related optimal weight) customised birthweight percentiles.

Contributions of authors

Tara TM Lee (https://orcid.org/0000-0001-6591-6421) (Obstetric Clinical Research Fellow, Norwich, PhD student, University of East Anglia) co-designed the study, recruited and provided support for trial participants and supported the preparation of results for the primary results publication and final report.

Corinne Collett (https://orcid.org/0000-0002-5510-1488) (Trial Manager, Norwich CTU) co-designed the study, supported study set-up procedures, data collection and analysis, including the statistical analyses, prepared the results for publication and co-wrote the primary results and final report.

Simon Bergford (https://orcid.org/0009-0004-0213-7059) (Statistician, Jaeb Center, USA) carried out data analysis, including the statistical analyses, and co-wrote the primary results paper and final report.

Sara Hartnell (https://orcid.org/0009-0002-8957-7924) (Diabetes Educator, Cambridge) provided participant and healthcare professional device training including troubleshooting and clinical care.

Eleanor M Scott (https://orcid.org/0000-0001-5395-8261) (Principal Investigator, Leeds) co-designed the study, screened and enrolled participants, provided patient care and/or took study samples.

Robert S Lindsay (https://orcid.org/0000-0002-9868-5217) (Principal Investigator, Glasgow) co-designed the study, screened and enrolled participants, provided patient care and/or took study samples.

Katharine F Hunt (https://orcid.org/0000-0002-1678-7835) (Principal Investigator, London) co-designed the study, screened and enrolled participants, provided patient care and/or took study samples.

David R. McCance (https://orcid.org/0000-0001-9491-8565) (Principal Investigator, Belfast) co-designed the study, screened and enrolled participants, provided patient care and/or took study samples.

Katharine Barnard-Kelly (https://orcid.org/0000-0002-3888-3123) (Professor of Health Psychology, Southampton) supported analysis of the patient-reported outcomes.

David Rankin (https://orcid.org/0000-0002-5835-3402) (Research Fellow Edinburgh) undertook and supported analysis of the qualitative interviews.

Julia Lawton (https://orcid.org/0000-0002-8016-7374) (Professor of Health and Social Science, Edinburgh) undertook and supported analysis of the qualitative interviews.

Rebecca M Reynolds (https://orcid.org/0000-0001-6226-8270) (Principal Investigator, Edinburgh) co-designed the study, screened and enrolled participants, provided patient care and/or took study samples.

Emma Flanagan (https://orcid.org/0000-0003-1492-7061) (Junior Trial Manager, Norwich CTU) supported the data collection, data analysis and preparation of results for the primary results publication and final report.

Matthew Hammond (https://orcid.org/0000-0002-0739-3412) (Director, Norwich CTU) co-designed the study and supported study set-up procedures.

Lee Shepstone (https://orcid.org/0000-0001-5524-7818) (Professor of Medical Statistics) co-designed the study and supported development of the statistical analysis plan.

Malgorzata E Wilinska (https://orcid.org/0000-0002-1564-8083) (Clinical Research Associate, Cambridge) supported the data collection and data analysis.

Judy Sibayan (https://orcid.org/0009-0005-9298-133X) (Principal Investigator, Jaeb Center, USA) co-designed the study and supported data analysis, including the statistical analyses, and preparation of results for the primary results publication and final report.

Craig Kollman (https://orcid.org/0009-0008-6696-0334) (Senior Statistician, Jaeb Center, USA) co-designed the study and carried out or supported data analysis, including the statistical analyses.

Roy Beck (https://orcid.org/0000-0002-5194-8446) (Director, Jaeb Center, USA) supported data analysis, including the statistical analyses, and preparation of results for the primary results publication and final report.

Roman Hovorka (https://orcid.org/0000-0003-2901-461X) (Professor of Metabolic Technology, Cambridge, UK) co-designed the study, designed and implemented the glucose controller and carried out or supported data analysis, including the statistical analyses and preparation of results for the primary results publication and final report.

Helen R Murphy (https://orcid.org/0000-0002-5489-0614) (Chief Investigator, Professor of Medicine, Norwich, UK) co-designed the study, screened and enrolled participants, provided patient care and/or took study samples, supported data analysis, including the statistical analyses, prepared the results for publication and co-wrote the primary results and final report.

All authors critically reviewed the report and contributed to the interpretation of the results. Tara TM Lee, Corinne Collett, Simon Bergford and Helen R Murphy are the guarantors of this work and, as such, had full access to the data and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors reviewed the paper prior to publication.

Our friend and collaborator Professor Fiona Denison (University of Edinburgh, Scotland) sadly died during the trial.

AiDAPT Collaborative Group members

Katharine Hunt, Helen Rogers, Damian Morris, Duncan Fowler, Josephine Rosier, Zeenat Banu, Sarah Barker, Gerry Rayman, Eleanor Gurnell, Caroline Byrne, Andrea Lake, Katy Davenport, Jeannie Grisoni, Shannon Savine, Helen R Murphy, Tara TM Lee, Tara Wallace, Alastair McKelvey, Nina Willer, Elizabeth Turner, Corinne Collett, Mei-See Man, Emma Flanagan, Matt Hammond, Lee Shepstone, Anna Brackenridge, Sara White, Anna Reid, Olanike Okolo, Eleanor M Scott, Del Endersby, Anna Dover, Frances Dougherty, Susan Johnston, Rebecca M Reynolds, Robert Lindsay, David Carty, Sharon Mackin, Isobel Crawford, Ross Buchan, David McCance, Louisa Jones, Joanne Quinn, Sarah Cains, Goher Ayman.

Disclosure of interests

Full disclosure of interests: Completed ICMJE forms for all authors, including all related interests, are available in the toolkit on the NIHR Journals Library report publication page at https://doi.org/10.3310/WCHZ4201.

Primary conflicts of interest: Sara Hartnell has been a member of the UK Medtronic Advisory Board 2020–22 and Dexcom Advisory Board 2021–22 and a director of Ask Diabetes Ltd providing training and research support in healthcare settings as well as reports having received speaker honoraria from Medtronic, Dexcom and Abbott Diabetes Care and consulting fees for CamDiab. Eleanor M Scott reports receiving speaker honoraria from Abbott Diabetes Care. Malgorzata E Wilinska reports patents related to closed-loop and is a consultant at CamDiab. Roman Hovorka reports receiving speaker honoraria from Eli Lilly, Dexcom and Novo Nordisk, receiving license and/or consultancy fees from B. Braun and Abbott Diabetes Care; holds patents related to closed-loop, and being is a director at CamDiab; and has advisory group roles for JDRF and Diabetes UK. Helen R. Murphy was a member of the HTA Maternal and Child Health Board 2016–9; is a current member of the UK and European Medtronic Scientific Advisory Boards and; reports speaker honoraria from Dexcom, Abbott Diabetes Care and Novo Nordisk; and has advisory group roles for JDRF and Diabetes UK. Katharine Barnard-Kelly is a recipient of a grant from Dexcom, is a Global Advisory board member for Roche Diabetes Care, is an ATTD conference speaker for Abbott Diabetes Care and is a co-founder and shareholder for Spotlight-AQ Ltd. Katharine F Hunt receives honoraria from the Association of British Clinical Diabetologists, Diabetes Technology Network and SPK Healthcare and is a member of Data Safety Monitoring Boards at the University of Cambridge. Craig Kollman is on the data safety monitoring board for INNODIA and reports being the recipient of JDRF grants and support from Tandem, Dexcom and Insulet made to the Jaeb institution. Tara TM Lee is the recipient of a Diabetes Research and Wellness Foundation Sutherland-Earl Clinical Fellowship. Roy Beck reports grants and equipment received to the Jaeb institution from NIH, JDRF, Helmseley Charitable Trust, Insulet, Tandem, Beta Bionics, Bigfoot Biomedical, Dexcom, Eli Lilly, and Novo Nordisk, as well as equipment support only from Ascencia, Medtronic, and Roche, as well as; consulting fees from Insulet, Tandem, Beta Bionics, Embecta, Vertex, Hagar, Ypsomed, and Zucara, he has received honoraria payments via the Jaeb institution for Tandem and the Endocrine Society. Judy Sibayan receives institutional grant funding and support from Jaeb for involvement in JDRF-funded studies. Julia Lawton was a member of the HTA General Committee 2018–9. Rebecca M Reynolds was a member of the NIHR HTA Commissioning Committee 2018–23 and is a current member of the NIHR HTA Funding Policy Committee (2021 onwards). David R. McCance, David Rankin, Simon Bergford, Emma Flanagan, Corinne Collett, Matthew Hammond, Lee Shepstone and Robert S Lindsay report no conflicts of interest.

Patient data statement

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it is important that there are safeguards to make sure that they are stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation

Data-sharing statement

De‐identified data sets will be made available on a case-by-case basis on reasonable request for research purposes. All requests should be addressed to the corresponding author.

Ethics statement

Approval was received from the East of England Research Ethics Committee (18/EE/0084) on 15 August 2018 and from the Medicines and Healthcare products Regulatory Agency.

Information governance statement

All information collected is held securely and treated in accordance with the Regulation (EU) 2016/679 (the ‘General Data Protection Regulation’ or ‘UK GDPR’) and the Data Protection Act 2018. University of East Anglia is the joint Data Controller for this trial with the Sponsor, Norfolk and Norwich University Hospitals NHS Foundation Trust. Similarly, UEA is the Data Processor for this trial. Martin Pond leads the Data Management Group for this trial. Norwich Clinical Trials Unit, Norwich Medical School, University of East Anglia, Norwich, UK. Email: martin.pond@uea.ac.uk Tel: 01603 591769.

Department of Health and Social Care disclaimer

This publication presents independent research commissioned by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, MRC, NIHR Coordinating Centre, the EME programme or the Department of Health and Social Care.

This monograph was published based on current knowledge at the time and date of publication. NIHR is committed to being inclusive and will continually monitor best practice and guidance in relation to terminology and language to ensure that we remain relevant to our stakeholders.

Disclaimers

This manuscript presents independent research. The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the MRC, the EME programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the EME programme or the Department of Health and Social Care.