Synopsis

Programme background and aims

Successful long-term management of type 1 diabetes (T1D) requires sustained individual engagement in complex self-management behaviours to maintain blood glucose [glycated haemoglobin (HbA1c)] levels within an optimal/safe range. Evaluation of Dose Adjustment For Normal Eating (DAFNE) educational course outcomes has shown that sometimes blood glucose levels drift from an optimal level over time.1 Behavioural science frameworks may be useful in developing standard DAFNE to support complex behavioural maintenance in the longer term.

In the narrative review undertaken by Heller et al.,2 the importance of ‘therapeutic education’, including the principles of flexible intensive insulin therapy (FIIT), in successful self-management programmes was highlighted. Learning should be a partnership between the healthcare professional (HCP) and the person with diabetes, enabling the empowerment of the latter to manage the condition in daily life, without any negative impact on the quality of life (QoL).

Although structured education programmes applying FIIT principles lead to clinically significant decreases in HbA1c, increased hypoglycaemia awareness and improvements in QOL, these benefits are not always sustained. The review outlined the paucity of evidence from previous studies. However, the DAFNE and related research programmes provide important new knowledge/understanding via their extensive qualitative work, suggesting roles in therapeutic interventions for: integrated and ongoing structured long-term support; inclusion of behaviour change theory to support initial and ongoing learning and long-term self-management; and effective technology which can facilitate links between professionals and a person with diabetes.

The overall aim of DAFNEplus was to refine the existing DAFNE curriculum, embedding behavioural science frameworks in all aspects of the intervention, addressing ‘ongoing support’, and including technological support, to aid long-term maintenance of HbA1c within an acceptable clinical range with no negative effects on participant QoL. Figure 1 outlines the stages and interactions of the different components of the research programme.

FIGURE 1 Research pathway diagram. BCT, behaviour change technique; GP, general practitioner; PPI, patient and public involvement; PROMs, patient-reported outcome measures; UCL, University College London; WS, workstream. (Figure not reproduced here; see the original article.)

Summary of alterations to the programme’s original aims/design

There were no major alterations to the overall aims and design of the programme. Some changes were undertaken in relation to the COVID-19 pandemic and to the withdrawal of one site in the final stages of randomised controlled trial (RCT) set-up.

One intervention site withdrew in the early stages of the trial, resulting in re-estimation of the power of the trial. This is detailed within the RCT section of the report.

The primary outcome of HbA1c was planned to be collected at site and analysed by a central lab. As a result of the COVID-19 pandemic restrictions, participants were unable to attend their local trusts for blood tests for extended periods of time. HbA1c home sample kits were introduced for part of the trial, when participants were unable to attend their clinical site for blood samples to be taken. Thus, due to the pandemic, we included HbA1c results from the three methods (central and local labs and the home sample kits).

The qualitative research team made some changes to their research design and questions in reaction to the pandemic. A smaller number of longitudinal case studies were completed, with a shift of focus to 12-month participant interviews. Additional qualitative 24-month interviews were also undertaken.

Some sites unable to restart after COVID-19 shutdowns due to capacity issues and/or issues in recruiting to a face-to face course.

Some consented participants were either unable to start, or to complete courses due to site closures in March 2020. Some of these participants returned and completed their courses when sites were allowed to reopen to research.

Workstream 1: revising Dose Adjustment For Normal Eating and developing Dose Adjustment For Normal Eating plus

It is unclear to the extent to which existing structured T1D education programmes, such as DAFNE, target the multiple, specific behaviours required for optimal self-management of T1D, and the factors influencing these behaviours. To examine this, and to identify opportunities for enhancing the existing DAFNE intervention to become ‘DAFNEplus’, we applied a state-of-the-art behavioural science framework, behaviour change wheel (BCW) approach to intervention development,3 which includes the Capability, Opportunity, Motivation, Behaviour (COM-B) model,3 and the associated behaviour change technique (BCT) taxonomy.4

The specific aims of this workstream (WS) were to:

1. specify behaviours involved in T1D self-management and barriers and enablers to sustaining these behaviours
2. specify BCT’s included in the existing DAFNE intervention and examine whether they target the identified barriers/enablers
3. identify additional BCTs with the potential to address barriers/enablers not currently targeted
4. generate recommendations for revising the existing DAFNE curricula via stakeholder consultation and consensus
5. integrate recommendations into a coproduced DAFNEplus intervention, pilot and refine the intervention ahead of the trial.

Methods and results

This mixed-methods WS involved five sequential steps, broadly corresponding to the steps of the BCW intervention development process (Report Supplementary Material 3, table 1).5 Engagement with adults with T1D and other multidisciplinary stakeholders throughout all steps was central to coproducing Dose Adjustment For Normal Eating plus (DAFNEplus). Full methods and findings have been published and reported elsewhere.5 Below, we summarise the methods and findings for each step in turn to illustrate the sequential intervention development process.

Ethical approval for intervention development was granted by University College London Research Ethics Committee (CEHP/2016/556) and for piloting by the NHS Research Ethics Committee (16/WS/0230).

Step 1: specifying target behaviours

Behaviours underpinning T1D self-management were identified via an in-person workshop with 16 stakeholders, including adults with T1D, clinical diabetologists, specialist nurses and dietitians experienced in delivering DAFNE, health and clinical psychologists, and behavioural scientists. Stakeholders described actions that people with T1D need to perform to self-manage T1D based on their experience of T1D and their review of the existing DAFNE curriculum. Transcripts of audio-recorded discussions were analysed qualitatively, and reported behaviours were inductively synthesised into flow diagrams depicting various cycles of self-management behaviours. Stakeholders reviewed draft diagrams, which were subsequently refined based on feedback.6

Stakeholders identified 150 distinct behaviours, summarised into 3 cycles of self-management behaviours: routine, reactive and reflective. Each cycle contains behaviours contributing to self-regulation, such as monitoring (e.g. checking glucose levels), and problem-solving (e.g. identifying patterns and reasons for glucose being in or out of target range). The three behavioural cycles are available in Report Supplementary Material 3, figure 1 and are reported in detail elsewhere.6

Step 2: identifying barriers and enablers

Next, we explored barriers/enablers influencing behaviours identified in Step 1. Factors influencing T1D self-management were extracted from four sources:

1. published systematic review and meta-ethnography7
2. a list produced by seven experienced DAFNE educators of specific behaviours and situations that they believe participants may need additional support with before, during or after participating in DAFNE
3. a list of ‘frequently asked questions’ generated by adults with T1D (‘users’) following participation in a DAFNE course
4. from the participants of an online DAFNE forum who were asked to report which aspects of self-management they most struggled with post course.6

Barriers/enablers extracted from each data source were first deductively coded to the domains of the COM-B model3 they were judged to best represent. Similar data points coded to the same domain were then grouped together and a summary theme label was inductively generated. Themes were categorised as a barrier, enabler or mixed influence on T1D self-management and were coded to the corresponding behaviour(s) identified in Step 1.

In total, 34 barriers and 5 enablers were identified: 9 within Capability (e.g. Psychological – forgetting glucose targets; Physical – impaired awareness of hypoglycaemia symptoms); 7 within Opportunity (e.g. Physical – lack of access to glucose monitoring equipment; Social – negative comparison with others’ progress); and 18 within Motivation (e.g. Automatic – feeling of failure and hopelessness; Reflective – perceived need for and/or consequences of glucose monitoring).6 All 39 themes are available in Report Supplementary Material 3, table 2.

Step 3: identifying additional behaviour change techniques

Next, we explored whether the BCTs identified as already included in the existing DAFNE curriculum were addressing the barriers/enablers identified in Step 2. We conducted a content analysis of the existing DAFNE intervention materials (i.e. curriculum manual, participant workbook and diaries), applying the BCT Taxonomy4 as a coding framework to identify and categorise component BCTs. We identified 24 unique BCTs across intervention materials (Report Supplementary Material 3, table 3). The most frequently used BCT was ‘Provide information about health consequences’ (e.g. information on insulin timing, leading to reduced risk of severe hypoglycaemia) (n = 17).5

We then triangulated findings from Steps 2 and 3 and consulted published matrices,8 which pair BCTs with COM-B, to suggest which BCTs are more likely to address barriers/enablers within particular COM-B domains. This enabled us to compare whether the BCTs in existing DAFNE materials are fully or partially addressing barriers/enablers identified in Step 2 (i.e. areas of match/mismatch between a BCT and behavioural influence), opportunities for refining how BCTs are delivered to better address barriers/enablers and missed opportunities (i.e. barriers/enablers identified in Step 2 not targeted by existing BCTs). Report Supplementary Material 3, table 4 presents identified barriers/enablers triangulated with BCTs.5 This resulted in a long list of potential BCTs for inclusion in DAFNEplus, which were discussed with multidisciplinary stakeholders. Each BCT was considered against the Acceptability, Practicability, Effectiveness, Affordability, Safety/Side-effects, Equity (APEASE) criteria,9 to include those potentially feasible and acceptable in DAFNEplus.

Step 4: stakeholder critical review and recommendations

In parallel to Step 3, we conducted another stakeholder consultation, asking 30 multidisciplinary stakeholders to review the existing DAFNE intervention content and generate recommendations for potential revisions to intervention content or mode of delivery.5 Stakeholders posted their recommendations on an online forum. In total, 474 comments were posted, which were inductively synthesised into 9 threads for further discussion under 3 categories: content (e.g. skills to develop), principles of programme delivery (e.g. terminology) and delivery format (e.g. layering content). These were posted on the online forum for further stakeholder comment. A face-to-face meeting was then held, where each recommendation was discussed until consensus was reached on its inclusion in DAFNEplus. Fifteen recommendations were agreed for inclusion (Report Supplementary Material 3, table 5).

Step 5: integrating, coproducing and piloting Dose Adjustment For Normal Eating plus

Finally, a multidisciplinary team of behavioural scientists, clinical diabetes experts (including diabetologists, specialist nurses, dietitians, clinical and health psychologists), engineers and adults with T1D integrated findings from Steps 1 to 4, and from our previous research,10–12 to produce the content, materials and guidelines for delivering DAFNEplus.5

A first iteration of the new DAFNEplus intervention was then piloted in three hospitals. In-person group sessions were observed by two researchers to explore the feasibility and fidelity of delivery. DAFNEplus participants and facilitators were interviewed following the course to explore acceptability, perceived value and ease of delivery. Observations and interviews identified recommendations for minor refinements to programme structure and materials (i.e. reordering sessions). Intervention materials were iteratively refined accordingly, with any refinements agreed collaboratively by the multidisciplinary research group.

The Dose Adjustment For Normal Eating plus intervention

Report Supplementary Material 3, table 6 presents the DAFNEplus intervention, specified according to the Template for Intervention Description and Replication reporting checklist.13 The final coproduced DAFNEplus intervention included three broad components:

1. a skills-based learning course comprising of 39 sessions delivered 1 day a week over 5 weeks, in groups of 6–8 participants, by 2 trained facilitators (nurse and dietitian)
2. an individual pre-course appointment and up to five structured individual follow-up support sessions, offered pro-actively by the facilitators over a 12-month period
3. an e-learning website and corresponding digital tool (the Glucollector) to support electronic logging, review and interpretation of data (e.g. on glucose levels, insulin dose and other factors), and communication between participants and HCPs.

These 3 components included 51 unique BCTs, of which 27 BCTs were new and additional to those included in the existing DAFNE materials.5 Report Supplementary Material 3, table 7 maps DAFNEplus BCTs to barriers/enablers identified in Step 2. Report Supplementary Material 3, table 8 compares BCTs in DAFNE versus DAFNEplus.

Strengths and limitations

Strengths included the triangulation of multiple data sources and stakeholder perspectives to inform intervention development. The application of behavioural science frameworks provided a structured, systematic, transparent, theoretically informed and evidence-based approach for refining existing complex behaviour change interventions. This enabled subsequent proposed mechanisms of action to be explored in a process evaluation. Nonetheless, this approach was resource-intensive. Challenges included balancing stakeholder perspectives and recommendations alongside theory-based approaches, which were mitigated through regular dialogue and consensus discussions.

Conclusions

This WS applied behavioural science to refine the existing DAFNE structured education programme into DAFNEplus. We identified underlying behaviours and influencing factors contributing to T1D self-management, and we identified 51 unique BCTs to address these. The DAFNEplus intervention was piloted and taken forward for evaluation of effectiveness and cost-effectiveness in a cluster-RCT (WS4), and a parallel process evaluation was performed to explore the mechanisms of action, acceptability and fidelity (WS4).

Workstream 2: structured support for ongoing diabetes self-management

Aims

The aims were to explore participants’ self-management after attending structured T1D education and their unmet need(s) for maintenance support to sustain improvements in glycaemic management and then to develop an evidence-informed model to guide effective follow-up support following DAFNEplus.

Methods and results

We undertook two approaches. The first examined qualitative studies which focused on self-management of T1D following structured education. The second was a systematic review of RCTs which evaluated self-management education, in particular those that provided ongoing structured support for at least 6 months following structured education.

(1) To inform the design of effective follow-up support, we undertook a meta-ethnography of qualitative studies focusing on self-management of T1D following structured education.

Thirteen RCTs delivering FIIT that had sufficient follow-up were analysed. Follow-up support delivered by the healthcare team providing the training, for at least 6 months, appeared to improve the effectiveness of the education programme based on comments recorded from both participants and educators. Additional features of support measures that can help include individual support (IS), use of technology with feedback, inclusion of cognitive behavioural therapy and involvement of the person’s support network. The design and delivery of follow-up support can influence the effectiveness of diabetes self-management education (DSME).

Noblit and Hare’s seven-stage meta-ethnography approach14 was used to offer new conceptual insights to inform the T1D evidence base and the design and evaluation of the ‘follow-up support’ aspect of DAFNEplus.

We expressed the synthesis as a logic model: Follow-Up Support for Effective T1D Self-management (FUSED) and the paper was published in 2018.7 It included 18 papers, published between 2010 and 2017, and these are listed in Appendix 2, Table 3. The FUSED model provided an explanation of why people find it difficult to maintain T1D self-management following structured education and outlined the elements required for effective follow-up support.

Following on from this activity, we undertook a further systematic review (as yet unpublished, but further details are provided below) to identify the most effective type of education programme to support sustained improvements in glycaemia.

We included studies evaluating a DSME programme delivered to participants with T1D, which incorporated training in FIIT. Where there were sufficient homogenous studies, data were pooled in meta-analyses. Where there was a lack of homogeneity in the included studies, a narrative synthesis was undertaken.

The model of structured follow-up support was led by a clinical psychologist (NdZ) who works closely with people with diabetes, and professionals delivering diabetes are with support from a second health psychologist. They worked closely with a multidisciplinary expert user group of clinicians to incorporate clinical psychological principles into the revised curriculum and in developing a new model of structured support.

This work linked into WS1 and the piloting section of the programme. Further information regarding development can be found there.

Strengths and limitations

The main limitations were that the review almost entirely included papers related to individuals who had participated in DAFNE courses in the UK and Ireland and therefore may not be applicable where other approaches are used. Also, while data from 18 publications were analysed, these came from only 6 unique studies.

Important limitations in the available evidence included: the paucity of studies of sufficient duration of follow-up measurements to allow exploration of the effects on HbA1c, limited descriptions of the follow-up support, risks to internal validity due to lack of assessor blinding, high attrition rates or poorly reported allocation concealment.

Conclusions

Structured education for T1D should offer an ongoing programme of individual follow-up support provided by specialist HCPs, which utilises technology and builds confidence in long-term self-management. This review provided some evidence to indicate key features to be considered in the development of follow-up support to facilitate sustained improvement of HbA1c, thus offering an evidence-based framework upon which to consider aspects of the developing DAFNEplus curriculum. This study provided a high-level logic model upon which to base follow-up requirements for the trial and added to the work undertaken in WS1 in identifying influences on self-management behaviours.

(2) We undertook a systematic review of follow-up and adjunct interventions to support self-management education for adults with T1D.

Methods and results

We undertook a systematic review of relevant RCTs, identified following a systematic search of electronic databases between 1 July 2016 and 28 April 2017. We identified three trials of DSMEs with supplementary educational approaches.15–17 All three trials significantly improved HbA1c levels, although in the one which compared structured group follow-up (Dinneen),15 there was no additional benefit in improving HbA1c. We identified one study investigating a programme for diabetes education and treatment for a self-determined living with type 1 diabetes efficacy of a self management programme (PRIMAS)16 which focused on the empowerment of people with T1D, enabling them to make informed judgements and choices about their own diabetes care. A key element was individual goal setting, and participants discussed the goals which they wanted to achieve within the course and reflected the extent to which they were able to achieve these goals and which barriers had to be overcome for an optimal goal attainment. There was a significant but modest reduction of HbA1c in participants who took part in PRIMAS compared with participants who took part in DTTP (change from baseline was −0.4%).

These results indicated that structured support in addition to technology (Glucollector), permitted participants to reflect on their glucose profiles (rather than rely on glucose readings collected on a meter) and reinforced important self-care behaviours (SCB). It also suggested that group-based support did not appear to have greater benefits over support provided in the usual individual format.

Strengths and limitations

There are important limitations in the evidence available. Very few studies have sufficient duration of follow-up measurements of both the intervention and control groups to enable us to explore the effects on HbA1c. There was also often a very limited description of what the follow-up support entailed – limiting our scope to compare follow-up interventions. Internal validity was frequently threated by lack of blinding of the assessor, high attrition rates, inadequate or poorly reported approaches to allocation concealment.

Conclusions

There is limited evidence from the RCTs that we analysed which would support the following features of follow-up to support sustained improvements in HbA1c:

• educational interventions that are followed up by structured support and contact from the specialist diabetes team who delivered the education
• support for a minimum of 6 months
• support incorporating both group and individual follow-ups
• support that involves an individual’s social support network.

Final remarks

The above findings led us to conclude that, in undertaking the DAFNEplus project, we should include structured support in addition to technology (Glucollector), permitting participants to reflect on their glucose profiles (rather than rely on glucose readings collected on a meter or book) plus addressing important SCB. Group-based support did not appear to have greater benefits over support provided in the usual individual format.

Workstream 3: using technology to support type 1 diabetes self-management behaviours

There were two main strands to the DAFNEplus technology developed for the programme:

A. Prototype Withcare+ and Glucollector systems

The Glucollector system integrates functionalities for seamless data acquisition, analysis and visualisation, along with intelligent decision support through e-learning, personalised communication and feedback to facilitate behaviour change and sustained T1D self-management.

At the core of the implementation of the Glucollector is the digital mapping of the routine, reactive and reflective cycles of T1D behaviours:5

• The reflective cycle is mapped onto functionalities supporting planning, implementing, reviewing, monitoring and problem-solving via visualisation, automated decision support and secure access to healthcare information and professionals (messaging).
• The routine cycle is mapped onto functionalities supporting ease of recording of diabetes related data, seamless data transfer and easy to check goal achievements.
• The reactive cycle is mapped onto functionalities supporting illness management, context enrichment of data with comments/data tags and knowledge deficiency.

The development of the Glucollector platform involved multiple iterations and active stakeholder involvement. Initial interviews with healthcare providers and patients identified potential improvements, focusing on workflows, patient–clinician engagement and technological needs for enhanced diabetes care.

An agile design methodology was adopted, involving adults with T1D and clinical experts in diabetes (doctors, dietitians, nurses and psychologists). High-risk, high-priority deliverables were addressed early, with the supervisory team ensuring compliance with medical regulations. Development cycles lasted for 2 weeks, with feedback from clinicians and participants integrated into sprints. Before deployment, the Glucollector was tested in four outpatient clinics for reliability and usability. Incubation cycles of 2–5 days allowed for rapid feedback and updates. Initially, only clinicians participated in these cycles for safety, but later, experienced DAFNE graduates and DAFNEplus cohort members were included. Features were piloted for up to 1 week before being made available to all users.

The study engaged three cohorts to evaluate the Glucollector system: University of Sheffield (UoS) affiliates, DAFNE graduates at Sheffield Teaching Hospitals (STH) and participants from the DAFNEplus multicentre pilot clinical trial.

Phase 1 – UoS volunteers (January 2016): 11 UoS students and staff with T1D piloted the Glucollector for 2 weeks, providing feedback through the integrated online system and completing a usability questionnaire.

Phase 2 – DAFNE graduates at STH (October 2016): 12 participants and 5 clinicians tested the system at home and in clinics, offering insights on its functions.

B. Prototype web-based educational resources

In the initial stages of the development, LifeGuide (https://lifeguideonline.org/) software was used to create interactive online modules specific to people with T1D, which could be used to support self-management by consolidating learning and promoting behaviour change during the course and follow-up period. The content included a recap of core DAFNE principles and interactive tools which incorporated BCTs, such as action planning and relapse prevention, to sustain self-management.

A user group of patient representatives, HCPs and psychologists collaborating on the DAFNEplus project provided informal feedback on the design and content of prototype LifeGuide web pages.

Outcomes

Withcare+ and Glucollector systems

The Glucollector is supported by the WithCare+ device for data acquisition and by developed bespoke tools for modelling, decision support and visual representation and interpretation.18–23 It incorporates functionalities to support shared visualisation for patients and clinicians based on collected data from various monitoring devices, automated feedback, goal setting and self-care facilitation. The WithCare+ device transmits data securely on SCB like blood glucose measurements [including both capillary glucose readings and continuous glucose monitoring (CGM) from Libre flash monitoring], carbohydrate intake and insulin doses. It supports linking to activity tracking devices and prompts users to contextualise abnormal readings. Clinician versions are available for in-clinic use.

The key features of the bespoke tools developed include:

• colour coding: highlights suboptimal glucose readings with medical device warning colours
• timely progress and trends: shows percentage of readings in range and estimated HbA1c
• pattern recognition: offers various data formats to identify significant patterns
• secure messaging: allows users and HCPs to discuss data and make adjustments
• amber flags: sends notifications to users and HCPs to highlight that action may need to be taken to support optimal management
• e-learning: provides modules on diabetes management topics
• gamification: encourages engagement through challenges, visual feedback and rewards
• caseload management: helps HCPs to prioritise follow up support
• tracking bolus advisor settings: records changes to insulin settings
• user feedback: allows users to report issues and suggest enhancements.

Prototype web-based educational resources

Most participants who accessed the e-learning valued the e-learning resources that could be accessed as required. Interviews provided feedback for improving the intervention and led to changes being made to adapt e-resources to those required by participants, reinforcing knowledge gained in the course. However, the LifeGuide components were not implemented into the final website design. Relatively few of those participating in the piloting accessed these resources, since only a limited number had access to the online material. In view of the increasing availability of online learning to those with internet access, and the limited funding (which precluded updating of these resources), we made the decision to focus the funding for technology on the Glucollector resource.

Pilot study

We undertook a pilot study to assess the feasibility and best approach to developing the logistics of the cRCT. We considered recruitment, intervention delivery, participant engagement and data collection methods, with the intention of refining the prototype DAFNEplus programme to optimise the intervention prior to the trial.

Research approvals for the pilot study were received from the West of Scotland Research Ethics Committee (ref16/WS/0230).

The prototype DAFNEplus programme was piloted in two waves at three DAFNE centres:

• Sheffield Teaching Hospitals NHS Foundation Trust
• King’s College Hospital NHS Foundation Trust
• Norfolk and Norwich University Hospitals NHS Foundation Trust

This was a non-randomised feasibility pilot where all participants received the intervention.

Inclusion criteria: Adults (≥ 18 years) with T1D for at least 6 months (or post ‘honeymoon’); prepared to undertake multiple daily injection (MDI) therapy and frequent self-monitoring of blood glucose; able to attend the full course 1 day per week over 5 weeks and participate in the follow-up.

Exclusion criteria: Use of insulin pump therapy; investigators used their judgement, informed by standard DAFNE guidelines, to determine whether the pilot study was suitable for potential participants who had: HbA1c > 12%; are pregnant; serious diabetes-related complications (e.g. blindness and renal dialysis) or other serious comorbidities; eating disorders; were unable to hear/speak/understand/read English; or were unable to give written informed consent.

The pilot intervention was delivered by practising DAFNE educators (diabetes specialist nurses, dietitians and physicians) at each participating site who had undertaken training in DAFNEplus.

The intervention protocol was described in a detailed DAFNEplus curriculum and programme manual. A clinical psychologist, specialised in the psychological needs of adults with diabetes and experienced in training diabetes professionals in behaviour change and communication skills, lead the development and delivery of educator training, with support from health psychologists from the University College London (UCL). Staff who completed the training programme were titled DAFNEplus ‘facilitators’, rather than educators, to emphasise the change in their method of delivering the intervention and supporting participants to develop their T1D self-management skills. The training process is described in the following section (see Training Dose Adjustment For Normal Eating plus facilitators).

No formal sample size calculation was required, as this was an assessment of the feasibility and acceptability of implementing the proposed procedures and processes for the trial.

We anticipated that the DAFNEplus programme could be optimised with participants from eight courses across the three centres. The number of courses was selected to enable all centres to complete at least two DAFNEplus courses during the pilot. The study was completed in two waves of iterative cycles to give centres experience in delivering the prototype programme and allow the study team the opportunity to refine the intervention based on feedback from educators, our patient and public involvement (PPI) group and findings from the process evaluation.

Participants on DAFNE waiting lists at each participating centre were invited to take part and asked to provide consent.

Eight DAFNEplus courses involving 54 patient participants were completed.

The collaborative working group

We developed a novel structured collaborative working group (CWG)7,24,25 process to use findings from the embedded qualitative work and feedback from other stakeholders (e.g. individuals involved in designing and delivering the intervention) to make iterative refinements to the intervention prior to the RCT.

Pilot study components

Pilot study process evaluation

Data collection methods included: Individual semistructured interviews with participants from the three pilot sites in Wave 1 took place at three time points: approximately 1 week after the course and 3 and 6 months later. We also interviewed Wave 2 participants post course and at 3 months.

Facilitator interviews and fidelity: Facilitators completed self-assessment checklists following every day of the DAFNEplus courses to assess intervention. The self-assessment checklist was piloted during course 1, and insights used to help refine documentation before subsequent courses. After completion of courses 1–3 and 6, facilitators were interviewed about their experiences of delivering the course (see WS1).

Observation: Researchers observed each day of courses 1–3 and course 6 during which they took detailed notes on course content and delivery. Each day of these courses was audio-recorded. All intervention sessions in courses 5–8 were audio-recorded with participant and educator consent. The purpose was to understand how the course was delivered and to develop the cRCT fidelity assessment.

DAFNEplus facilitator interviews: Staff were interviewed about their experiences of attending the DAFNEplus training sessions and delivering the intervention.

Data analysis was informed by thematic and framework approaches26 and was supported through use of NVivo software (QSR International, Warrington, UK).

A number of refinements were made to the intervention (course, IS sessions and Glucollector) in light of feedback from participants and input from the CWG. See Breckenridge et al.24 for some illustrative examples.

Glucollector/Withcare+ user testing

We incorporated a parallel phase of user testing of the Glucollector/Withcare+ system. Qualitative data were gathered via the process evaluation on participant and educator views and experiences of using the system (see Pilot study process evaluation).

Additionally, study participants were asked to complete the System Usability Scale to provide feedback on using the technology at 6 months. The Glucollector website included an in-built online feedback mechanism to enable participants to give feedback as and when they saw fit.

Participant and educator data on system usage were collated, for example, frequency of site access, average length of time per session, number of pages accessed and popularity of different pages. Numerical data on mean blood glucose and percentage of readings in the hypoglycaemic range were also examined.

Collection of outcome measures

We collected the clinical, psychological and health economic outcomes to inform the feasibility of collection in the cRCT. A list of the outcome measure completed during the pilot study can be seen in Appendix 3.

Given the early lack of consensus about the optimal measure of diabetes-specific QoL to be used as the primary psychological outcome for this RCT, we carried out a large-scale online survey of adults with T1D to determine the acceptability, relevance and comprehensiveness of five diabetes-specific QoL measures (see Appendix 1). The findings supported use of the Audit of Diabetes-Dependent Quality of Life (ADDQoL) as the primary psychological outcome, and the Diabetes Attitudes Wishes and Needs Impact of Diabetes Profile (DIDP) as a secondary outcome to determine its responsiveness for future trials.27

Psychological measures were selected according to existing knowledge about their association with the trial’s primary outcome (HbA1c) and diabetes-specific QoL (primary psychological outcome), the results of the ‘Your Self-management And You: Quality of Life’ (YourSAY:QoL) survey (see Appendix 1), our former work with the DAFNE intervention, and based on the theoretical framework that underpins the new intervention development work and possible treatment mechanisms.27

Clinical data were collected directly from the participants, or via DAFNE educators or other HCPs at sites, via the Glucollector or LifeGuide websites, or source documents (e.g. clinical notes), and were input onto case report forms. Questionnaires, including the psychological and health economic outcome measures, were completed by participants on paper at baseline and 6-month study visits, or were posted to non-attenders.

As the purpose of collecting the clinical and psychological outcomes was to assess the feasibility of data collection methods and likely completion rates by participants, with a view to determining acceptability and participant burden, analysis was descriptive and did not include formal hypothesis testing.

Pilot study key outcomes

We piloted and iteratively amended the DAFNEplus curriculum via two waves of pilot courses. The qualitative research findings, practitioner reflections and PPI group feedback were iteratively developed by the CWG into an updated version of the curriculum, which was delivered during the interim courses and finalised prior to the cRCT. All the supporting resources for the course were also prepared.

All eight DAFNEplus cohorts in the pilot study were given access to the technology and structured follow-up support as part of the intervention. In line with the work developing and refining the curriculum, the CWG used the same approach to optimising the structured support and technology.

The qualitative research with DAFNEplus participants and facilitators enabled the identification of key barriers and facilitators to the delivery and receipt of the intervention in real time. In addition, we were able, through the CWG, to mobilise the knowledge generated through both the pilot study and PPI feedback to iteratively develop the intervention and address any identified shortcomings.

The YourSAY study advised on the primary and other psychological-related patient-reported outcome measures.

The health economists completed an evaluation of the potential cost-effectiveness of DAFNEplus compared to standard DAFNE in an NHS setting. This is detailed within the Health Economic evaluation.

Training Dose Adjustment For Normal Eating plus facilitators

Training of the facilitators included three parts: self-directed learning (2 days), a live workshop (3 days) and a 2-hour webinar (Figure 2). Led by clinical psychologist (NdZ), the training team included a clinical psychologist (PW), a diabetes specialist nurse (CF) and a diabetes specialist dietician (CG), who were all involved in the training design and delivery.

FIGURE 2 Overview of facilitator training. (Figure not reproduced here; see the original article.)

Facilitators first completed three self-directed learning modules that provided the background, theory and context for DAFNEplus. They then attended a 3-day, in-person, training workshop, which provided interactive and experiential learning. Training methods included: self-directed learning, didactic teaching, small group discussion and role play. The learning followed an iterative process of observation, reflection, practice and feedback. The training philosophy mirrored the guidance given to DAFNEplus participants: ‘there is no failure, only feedback’. Consequently, the focus was upon trying new things (e.g. role plays) and the benefits of making mistakes in a supportive environment.

During the cRCT, facilitators received ongoing support. For the first DAFNEplus group, facilitators had a support phone call before, and e-mail supervision after, each week of the course. From the second DAFNEplus group onwards, facilitators received weekly e-mail supervision, and ad hoc e-mail/phone support was also available.

After delivering the first DAFNEplus group, facilitators attended a 2-hour webinar [plus live Questions & Answers (Q&A)] to train in the IS component of the DAFNEplus programme. In total, 28 facilitators were trained from 6 DAFNE centres.

Detail of the above process has recently been published in de Zoysa et al.28

Workstream 4: cluster randomised controlled trial with concurrent process and health economic evaluations

Introduction

A cRCT was conducted to examine the effectiveness of the intervention (DAFNEplus) compared with standard DAFNE. Concurrent health economic and process evaluations were undertaken. The process evaluation included four interlinking components: (1) interviews, (2) psychological process measures, (3) technology measures and (4) fidelity.

The aim of this superiority cRCT was to investigate whether DAFNEplus enabled improved and sustained diabetes self-management behaviours, leading to better glucose management (i.e. reduced mean HbA1c by minimum clinically significant difference of 0.5%) at 12 months than achieved with standard DAFNE, without compromising the QoL.29

Objectives

The primary objective was to compare the effect of DAFNEplus to standard DAFNE on glycaemia (HbA1c) at 12 months (primary biomedical outcome).

The secondary objectives were:

• to compare the effects of DAFNEplus to standard DAFNE on:
  • diabetes-specific QoL at 12 months (primary psychological outcome)
  • glycaemia (HbA1c) at 6 months (medium-term outcome)
  • other biomedical outcomes, including severe hypoglycaemic episodes, diabetic ketoacidosis (DKA), weight, body mass index, blood pressure and lipids
  • secondary psychological outcomes, including diabetes distress, diabetes-specific positive well-being and fear of hypoglycaemia
• to undertake a mixed-methods process evaluation to aid understanding of the cRCT findings and inform decision-making about the implementation of DAFNEplus in clinical care post trial
• to assess the fidelity of delivery of DAFNEplus
• to undertake a health economic analysis to determine the cost-effectiveness of DAFNEplus versus standard DAFNE.

Design

The trial used a pragmatic, parallel group, superiority cluster randomised (1 : 1 allocation) controlled design involving 14 NHS specialist diabetes centres. Centre randomisation was implemented rather than individual participants, as contamination of the control arm may have occurred if educators were trained in DAFNEplus30 and still were required to deliver standard DAFNE (control).

Cluster randomised controlled trial methods

Sample size

It was expected that there would be 882 people with T1D referred for DAFNE courses within the recruitment window and that, of these, 75% (662) would be recruited, equivalent to 47 participants at each of the 14 centres. Based on data from previous DAFNE courses, a further 25% were expected not to meet the primary analysis population criteria of a baseline HbA1c > 7.5%, leaving 497 participants. Finally, we anticipated 15% of participants to be lost to follow-up by the 12-month stage, therefore giving a primary analysis population of 422 participants.

Considering a design effect, due to the cluster design of the trial, with an intraclass correlation coefficient of 1.5% (from previous DAFNE data) and 30 participants per cluster (422 over 14 centres), the design effect was 1.435, leaving the effective total sample size of 294 participants (147 per arm). Using a two-sample comparison of mean HbA1c at the 12-month follow-up with 2-sided alpha of 5%, a correlation of 0.5 between baseline and final values and a standard deviation of 1.45 (from previous DAFNE data), the trial sample gives 92% power to detect a 0.5% difference in HbA1c between the two treatment groups.

The original calculations gave a power of 92.7%. During the study, one of the intervention centres withdrew from the trial. Due to the challenges of enlisting new centres at that stage, the Trial Steering Committee agreed that the trial could continue with 13 instead of 14 centres (6 in the intervention, and 7 in the control arm) and the original planned recruitment per centre. This approach reduced the sample size solely in the intervention arm and provided 90.4% power with a small imbalance between the two arms (ratio 1 : 1.167), with a reduced sample size of 615.

Participants

Site level

Study centres (sites) were considered to be eligible if they met the following criteria:

1. adult T1D centre delivering DAFNE
2. at least three DAFNE educators trained to deliver the 5-week model of DAFNE
3. delivery of sufficient DAFNE courses per year to recruit study sample.

Individual participants

Potential participants were identified by local diabetes clinicians using standard criteria for referral to DAFNE.

Participants were considered to be eligible if they were ≥ 18 years of age, had a diagnosis of T1D for at least 6 months (or post honeymoon), were prepared to undertake MDI therapy and frequent self-monitoring of blood glucose, could attend all sessions as part of the intervention and the principal investigator was confident that they could adhere to the protocol requirements. Please see Figure 3 for the flow of participants through the trial.

FIGURE 3 The DAFNEplus Consolidated Standards of Reporting Trials diagram. (Figure not reproduced here; see the original article.)

Those currently using continuous subcutaneous (sc) insulin infusion pump therapy or who had attended a DAFNE course within the last 5 years were ineligible. Where potential participants had HbA1c > 12%/108 mmol/mol, serious diabetic complications (e.g. blindness and renal dialysis) and other serious comorbidities (e.g. psychosis and diagnosed eating disorder), investigators could use their judgement regarding inclusion, informed by standard DAFNE guidelines and in agreement with the trial team. Participants also needed to have a good understanding of spoken and written English and have the capacity to provide informed consent.

Intervention

DAFNEplus was facilitated by practising DAFNE educators (diabetes nurses, dietitians and physicians) in the NHS, all of whom had previous experience in DAFNE training. An extensive training package was undertaken by all intervention staff (see Training Dose Adjustment For Normal Eating plus facilitators).

DAFNEplus is a complex intervention, as described in the previous sections on development and piloting of the intervention.

Control arm

Centres allocated to the control arm delivered standard DAFNE courses. All centres delivered a 5-day course, 1 day a week over 5 weeks.

See Table 1 for a summary of the intervention and control programmes.

Randomisation

All participating centres were randomised on a 1 : 1 basis to control (standard DAFNE course) or the intervention arm (DAFNEplus programme). A covariate-constrained approach matched the centres on the number of patients, number of educators and the total number of previous DAFNE courses delivered by the centre as stratification variables.

Blinding

Due to the nature of the intervention, it was not possible for participants or members of the study team working directly with participants or the intervention to be blinded. Outcome assessors were also unblinded. The trial statisticians did not have access to any outcome data until the time of the analysis.

Outcome measures

The primary biomedical outcome was HbA1c at 12 months.

The study was powered for this single, primary end point. All additional analyses were performed to further inform this primary end point and should be interpreted accordingly.

The HbA1c at 6 months post course was a secondary outcome. Additionally, the number of participants achieving glycaemic control [defined as either a HbA1c < 7.5% (58 mmol/mol) or a decrease in HbA1c of ≥ 0.5% (≥ 5.5 mmol/mol)] were calculated at both 6 and 12 months post course. These cut-off points are recognised throughout the diabetes research community as being clinically relevant.

Other secondary biomedical outcomes included:

1. severe hypoglycaemia, as defined by the American Diabetes Association, both rates and proportion of those affected
2. DKA, both rates and proportion of those affected
3. weight
4. body mass index
5. blood pressure
6. lipids.

Participant safety was assessed by way of adverse and serious adverse reporting. All participants were included in safety summaries.

We also collected and will analyse 24-month outcome data (HbA1c and severe hypoglycaemic episodes) after the main study has been reported. See Appendix 3, Table 4 for a list of the biomedical outcomes.

The primary psychological outcome was diabetes-specific QoL at 12 months, measured using the ADDQoL-15, specifically the average weighted impact (AWI) score, assessed by summing the weighted impact of diabetes for each item (i.e. impact score multiplied by the importance score), divided by the number of applicable items. Secondary psychological outcomes included: diabetes-specific QoL (DIDP); diabetes distress [Problem Areas in Diabetes scale-11 (PAID)-11)]; diabetes-specific positive well-being [4-item subscale of the Well-Being Questionnaire (28 items) (W-BQ28)]; and fear of hypoglycaemia (HFS-II short-form). Scoring is provided in Appendix 3.

Statistical analysis

Three analyses populations were defined:

The primary intention-to-treat (ITT) population composed of consented participants with baseline HbA1c > 7.5%, analysed as per their randomisation allocation, regardless of treatment adherence [defined as per protocol (PP)] and protocol non-compliances.

Full ITT population composed of consented participants, analysed as per their randomisation allocation, regardless of treatment adherence and protocol non-compliances.

The PP population composed of the subset of the primary ITT population who adhered to their assigned intervention.

For all analysis populations, participants recruited in 2020, who did not attend at least one session of the initial DAFNE course, were excluded. This was due to courses not running during the COVID-19 pandemic.

The primary biomedical outcome, difference in HbA1c at 12 months between the groups in the primary ITT population, was analysed using generalised estimating equations, with clustering by centre, robust standard errors, adjusted for baseline HbA1c, sex, age at baseline, use of Libre/CGM at baseline and an exchangeable correlation structure. Multiple imputation was used to handle missing outcome data.

Full details of the planned biomedical analysis are reported in the statistical analysis plan (see Report Supplementary Material 4).

For the psychological outcomes, for the full ITT cohort, a mixed linear regression model using restricted maximum likelihood estimation was used to assess the mean difference in primary and secondary psychological outcomes between the study groups at 12 months. The outcomes at 6 and 12 months were included as dependent variables in the model. Time (baseline, 6 and 12 months), the outcome at baseline and an interaction between study group and time were included in the model as fixed effects. The model included random intercepts to accounting for clustering within individuals and centres, and an unstructured variance–covariance structure was assumed for the random effects. Age, sex and use of CGM at baseline were included as fixed effects in the adjusted model. The estimated mean differences, 95% confidence intervals (CIs), p-values and standardised effects sizes (Cohen’s d) were reported for both unadjusted and adjusted models. Transformations were considered for skewed outcomes and model diagnostics performed.

Two sensitivity analyses were performed for the primary psychological outcome. First, the robustness of the missing at random assumption of the mixed model was tested via pattern-mixture modelling. Results were compared with that obtained in the primary analysis. A PP analysis was then performed to estimate the intervention effect in those complying with the intervention as outlined in the protocol.

Randomised controlled trial results

Key findings: biomedical outcomes

In the primary ITT population, mean HbA1c reduced from 9.1% (standard deviation 1.4) to 8.2% (1.2) [75.9 (15.0) to 66.5 (13.0) mmol/mol] in the DAFNE group and from 9.1% (1.3) to 8.1% (1.1) [75.6 (14.1) to 65.1 (12.4) mmol/mol] in the DAFNEplus group.

There was no evidence of a between-arm difference in HbA1c at 12 months {point estimate [95% CI −0.014 (−0.284 to 0.255); p = 0.915]} in the primary ITT population. This was consistent with sensitivity analyses, including those adjusting for adherence, and varying assumptions about missing data.

In the full ITT population similarly, no evidence of between-group difference was found for any of the secondary end points.

The number of adverse events reported was low. During follow-up, at least one episode of severe hypoglycaemia was experienced by 3% in the DAFNEplus arm (vs. 7% in the DAFNE arm) in the full as-randomised population, with 1% of participants being hospitalised as a result of these episodes in each arm. In each arm, 3% experienced at least one episode of DKA during the follow-up in the full as-randomised population, with 1% in the DAFNE arm and 2% in the DAFNEplus arm being hospitalised as a result of these episodes.

Key findings: psychological outcomes

In the ITT analyses, the adjusted mean difference in the impact of diabetes on QoL [i.e. the ADDQoL-15 AWI score at 12 months (primary psychological outcome)] was 0.2 (95% CI −0.1 to 0.5; p = 0.21), equivalent to a standardised effect size of d= 0.1 (95% CI −0.1 to 0.4) and not statistically significant.

For the secondary psychological outcomes, at 12 months, there was a statistically significant difference in diabetes distress (PAID: adjusted mean difference −3.3, p = 0.001 and d = −0.4) and hypoglycaemia avoidance (HFS-II behaviour subscale: adjusted mean difference −1.1, p = 0.002 and d = −0.3) and diabetes-specific positive well-being (4-item subscale of the WBQ-28: adjusted mean difference 0.9, p = 0.004 and d = 0.4), favouring the DAFNEplus intervention. No statistically significant changes at 12 months were observed for the secondary diabetes-specific QoL (assessed with the DIDP: adjusted mean difference −0.1, p-value= 0.32 and d = −0.1) or hypoglycaemia worries (HFS worry subscale: adjusted mean difference −0.9, p = 0.05 and d = −0.2).

Few statistically significant between-arm differences in psychological process evaluation scores were observed at course completion, or at 9 months, thereby aligning with the primary outcomes. Exceptions included difference between arms in satisfaction with diabetes management [Diabetes Management Experience Questionnaire (DME-Q) subscales] at 9 months only, and certain SCB [self-care behaviours–T1D (SCB–T1D)], both favouring DAFNE.

Key findings: health economic outcomes

Our primary base case within-trial analysis using economic evaluation alongside the trial (EEACT) methods showed that DAFNEplus had an incremental cost of £611 versus DAFNE, an incremental quality-adjusted life-year (QALY) of + 0.008 and an incremental cost-effectiveness ratio (ICER) of £78,147 per QALY gained. All our scenario analyses, including complete-case analysis and conducting a long-term extrapolation model of biomedical outcomes, had ICERs > £20,000 per QALY gained. National Institute for Health and Care Excellence’s (NICE’s) typical range for a maximum acceptable ICER is £20,000–30,000 per QALY gained, and the ICER for DAFNEplus versus DAFNE is highly likely to exceed this range.

Limitations

Biomedical analysis

The primary analysis was subject to 31% of missing outcome data overall. This may have introduced bias into the results, although our sensitivity analysis indicates that results are robust to assumptions made about missing data. Clustering may have introduced some baseline differences.

The COVID-19 lockdowns resulted in participants being unable to attend clinics for HbA1c bloods collection. We moved to the use of self-testing kits which were provided and analysed by The Doctors Laboratory (TDL) (www.tdlpathology.com/) for the duration of lockdown. Participants’ use of CGM increased from 51% at baseline to 78% at 12 months. This may have contributed to the overall improvement in HbA1c scores.

Psychological analysis

For the primary psychological outcome, there was approximately 46% missing data in DAFNEplus arm and 40% in DAFNE arm at 12 months. The sensitivity analysis showed that the mean of the missing ADDQoL data would need to be at least 2 points higher than the mean of the observed ADDQoL data (in the DAFNEplus arm only) to see an intervention effect > 1, which the authors did not deem likely.

Another limitation is that numbers represent mean differences and thereby do not allow for the possibility that certain subgroups of participants may benefit more (or less) from one intervention over the other. Latent class or profile analyses may offer further insights.

Only statistically significant differences, as well as standardised effect sizes, were considered, as minimally important differences have not been established, either from a clinical or participant perspective. It is also important to note that the cRCT was powered to detect a difference in HbA1c, but it was not powered specifically for the psychological outcomes.

Process evaluation

The process evaluation is composed of four interlinking and complementary components: (1) qualitative interviews, (2) quantitative psychological process measures, (3) technology measures and (4) mixed-methods fidelity assessments.

Objectives

The overarching research questions for the process evaluation are:

1. Does the DAFNEplus intervention ‘work’ in the ways intended? If not, why not?
2. What are the implications of the findings of the process evaluation for the roll-out of DAFNEplus in routine clinical practice?

To answer these questions, a series of overlapping subquestions were explored and are included below. The data sources which were used to answer these are detailed in Appendix 4, Table 6.

Participants and consent

All trial participants consented to be contacted for interviews as part of the overall consent process. Those who attended sites due to be recorded for the fidelity aspects of the trial were also asked to consent to be recorded. Facilitators were consented into the qualitative and fidelity assessment work at the start of their involvement in the trial.

Further detail about sampling is included in each of the process evaluation components.

Component 1: interviews

Full details of the objectives and methods are provided in Lawton31,32 and Hart.33

Interviews with Dose Adjustment For Normal Eating plus participants

We interviewed n = 28 adults with T1D who attended a DAFNEplus course and received at least three individual follow-up sessions.

Interviews with Dose Adjustment For Normal Eating plus facilitators

We interviewed n = 18 facilitators (diabetes specialist nurses and dietitians) who were involved in delivering DAFNEplus.

Component 2: psychological process measures

As the primary biomedical and psychological outcomes of the cRCT showed no statistically significant between-group differences, the planned process evaluation (i.e. to identify mediating pathways of a between-arm difference in HbA1c and diabetes-specific QoL at 12 months) was abandoned. Instead, an exploratory analysis was conducted to understand psychological processes and their associations with primary outcomes (HbA1c and ADDQoL AWI score).

Methods

At baseline, course completion, 3 and 9 months post course, participants completed psychological process measures, assessing constructs such as diabetes-specific SCB, self-regulatory skills and satisfaction with diabetes management (see Appendix 3, Table 5).

Statistical analysis

Descriptive statistics were used to describe the psychological processes at all time points. Mixed linear regression models using restricted maximum likelihood estimation were applied to assess the difference in psychological process evaluation measures (dependent variable) over time (from baseline to course completion, course completion to 3 months and 3–9 months). Calculation of total and subscale scores is described in Appendix 3. Independent variables (fixed effects) included time (baseline, course completion, 3 and 9 months) and randomisation (DAFNE and DAFNEplus), as well as control variables sex (male and female), use of CGM (yes and no) and age. An interaction between randomisation and time was included to determine the potential between-arm differences. Random intercepts were used to account for clustering within individuals and centres. To explore associations between changes in process evaluation measures (from baseline to course completion) and changes in primary outcomes (HbA1c or ADDQoL AWI, from baseline to 12 months), standard linear regression models were used. Change scores in primary outcomes (HbA1c or ADDQoL AWI) were used as the dependent variables, while independent variables included baseline HbA1c or ADDQoL AWI score, change scores in process evaluation measures, randomisation, sex, use of CGM and age.

Component 4: fidelity

We explored the extent to which DAFNEplus was delivered as intended (with fidelity) by exploring:

1. To what extent was manual-specified content delivered as intended in DAFNEplus29 and DAFNE (comparator) sessions?
2. Did fidelity vary across sessions, over time and sites?
3. Were there any additions or modifications to content delivered, and why?

Design

Mixed-methods study, involving self-reported fidelity checklists and objectively verified fidelity through content analyses of audio-recorded intervention sessions, was conducted.34 We focused on the pre-course assessment, group sessions and IS sessions of DAFNEplus for this fidelity assessment (i.e. not digital components – e.g. Glucollector), as these components involved an interaction between the facilitators and course participants and thus had more scope for fidelity drift.

Fidelity checklists

All facilitators delivering both DAFNEplus and standard DAFNE sessions completed a self-reported fidelity checklist after each session (see Report Supplementary Material 5, table 1). Checklists were structured around the DAFNEplus/DAFNE manuals, which specified activities to be delivered during each session. As part of intervention development,5 each activity was mapped to corresponding BCTs.4 Each checklist included a table, with each row representing an intended activity, which facilitators rated as delivered ‘fully’ (2), ‘partially’ (1) or ‘not at all’ (0). A free-text box was available to specify reasons for not delivering the activity or describe modifications or additions to session content. To quantify the overall fidelity for each session, we calculated the percentage of manual specified activities fully delivered in each session. We also analysed fidelity by BCT, calculating the percentage of instances where each BCT was fully delivered as intended across sessions.34 According to published criteria, ≤ 50% delivery was classified as ‘low’, 51–79% as ‘moderate’ and 80–100% as ‘high’ fidelity.35 We compared overall fidelity percentage scores over time, across session types, sites and between DAFNEplus and DAFNE sessions. Free-text responses were analysed thematically. Themes representing reasons for not delivering activities were mapped to the COM-B model.8

Recorded sessions

Half of participating sites (n = 4 DAFNEplus, n = 2 DAFNE) also audio-recorded all group support sessions. A subsample of n = 27 sessions were purposively sampled for analysis according to whether they met either of the following criteria: (1) sessions with lowest self-reported fidelity of delivery from the checklists, to provide further insights into how and why fidelity might have been lower; and/or (2) one of each DAFNEplus sessions with more novel content relative to DAFNE, to explore how newer content was delivered (see Report Supplementary Material 3, table 8).5 Within these criteria, we sampled to maximise variation across sites and courses. The sessions included are listed in Report Supplementary Material 5, table 2.

We content-analysed session transcripts, focusing on the speech of facilitators to analyse what was delivered to participants. Two independent researchers used the corresponding fidelity checklists, described above as a coding frame, assessing whether each intended manual-specified activity for that session was delivered fully/partially/not at all. Inter-rater reliability was high (average 92% agreement; range: 82–100%). We also coded session transcripts into component BCTs using BCT Taxonomy v1,4 and we calculated the number of times the BCT was delivered in that session according to manual specification.30

Strengths and limitations

Given the large volume of sessions, sites and courses, it was not possible to objectively verify fidelity via a content analysis of all recorded sessions. Therefore, a strength is the mixed-methods approach, balancing breadth of data collection across sessions and sites with checklists, alongside in-depth analysis of a subsample of audio-recorded sessions. However, the discrepancy in fidelity results between checklists and recordings suggests that checklist data was prone to bias (i.e. self-report, social desirability and recall), with facilitators overestimating their fidelity of delivery. Facilitators knew they were being observed and assessed for fidelity, potentially leading to a Hawthorne effect.36 Therefore, the moderate fidelity levels observed, likely represent a ‘best case’ fidelity scenario, and potential actual fidelity may have been lower when not observed.37 Some caution is needed when interpreting fidelity % scores due to low frequency of some activities in some sessions (e.g. BCT delivered once, but intended frequency is two = 50%, low fidelity).

Process evaluation results

The findings for each of the research questions are summarised below, using data from one or more of the process evaluation components (interviews, psychological process measures, technology measures and fidelity). Note that not all components are relevant for all questions.

1. What changes in psychosocial process measures were observed from baseline to DAFNEplus course completion?

Overall, from baseline to course completion, there were statistically significant improvements in all psychological process measures (this excludes SCB; change over time was not assessed due to the extensive number of items and subscales) in both DAFNEplus and DAFNE arms (majority p < 0.001). These included satisfaction with diabetes management (DME-Q), self-regulatory skills [Self-Regulation Questionnaire for type 1 diabetes (SRQ–T1D)], Diabetes Strengths and Resilience Questionnaire (DSRQ), beliefs about consequences (BCQ) and beliefs about capabilities in self-managing diabetes [Confidence In Diabetes Scale (CIDS)] and staying safe from hypoglycaemia-related problems [Hypoglycaemia Confidence Scale (HCS)]. There were few additional changes occurring later, suggesting that improvements were maintained over time. Further changes related mainly to further improvements in satisfaction with diabetes management (DME-Q subscales) and worsening of BCQ.

1. What mechanisms of change impact glycaemic control among DAFNEplus participants?
   • Psychological process measures

Improvements in HbA1c (from baseline to 12 months) were associated with changes in the following psychological processes (from baseline to course completion): less convenience of diabetes management (DME-Q subscale) and improvements of certain SCB (SCB–T1D subscale: ‘monitoring and adjustment re. stress and weight’).

1. What mechanisms of change impact diabetes-specific QoL among DAFNEplus participants?
   • Psychological process measures

Reduction in the negative impact of diabetes on QoL, from baseline to 12 months, was generally not significantly associated with changes in psychological process measures (from baseline to course completion).

1. What are participants’ experiences of, and views about, key elements of DAFNEplus and how do these influence and inform changes in, and maintenance of, key diabetes self-management behaviours over time?

Participants’ experiences of, and engagement with, the course and IS components are described in full elsewhere.31,33

In brief, all participants, albeit to varying degrees, reported gaining benefit from the DAFNEplus programme. As a result of engaging with the programme’s novel content and BCTs, participants experienced changes in their mindsets and emotional resilience, which increased their confidence and motivation to apply DAFNEplus principles and skills in their everyday lives.31 These benefits were also reinforced by the IS sessions, which fostered further improvements in participants’: (1) FIIT knowledge and data interpretation/pattern spotting skills (‘capability’); (2) confidence, autonomy and ability to access timely support from appropriate HCPs when needed (‘opportunity’) and (3) perceptions of the value of enacting FIIT principles in everyday life (‘motivation’). Increasing independence in self-management was also encouraged and enabled by the increasing spacing of the IS sessions over the 1 year of follow-up.33

These findings suggest that DAFNEplus worked as intended, that is consistent with the programme theory (see Report Supplementary Material 1).38 However, many participants also highlighted a need for ongoing, tailored support from DAFNEplus trained HCPs to further consolidate and/or retain programme benefits over time.

1. To what extent is the DAFNEplus intervention delivered as intended?
   • Fidelity assessments

We received 2191 checklists (n = 1244 for DAFNEplus, n = 946 for DAFNE sessions) (see Report Supplementary Material 5, table 3). Overall, self-reported fidelity was high: on average, 90% of manual-specified activities were self-reported as delivered in each DAFNEplus session (range: 64–97%). Fidelity was highest for sessions: ‘monitoring and long-term health’, ‘lapse/relapse management and management plans’ and ‘looking back/looking forward- maintaining DAFNEplus’ (99%) (see Report Supplementary Material 5, table 4). Fidelity was lower for IS sessions (70–72%), carbohydrate counting in real settings (68%) and Q&A with doctors (64%). Each BCT was, on average, delivered in 86% of intended instances (range: 42–100%), representing high fidelity. Fidelity was lower for BCTs: ‘social-incentive’ (n = 42%) and ‘non-specific reward’ (n = 61%) (see Report Supplementary Material 5, table 5). There was no discernible pattern in self-reported fidelity across sites or over time (see Report Supplementary Material 5, figure 1).

Overall, self-reported fidelity scores were also high in standard DAFNE sessions (average 87%, range: 72–96%; see Report Supplementary Material 5, table 4).

Objectively verified fidelity in session transcripts was consistently lower than self-reported fidelity (see Report Supplementary Material 5, figure 2). On average, 59% (range: 31–90%) of manual-specified activities were fully delivered in the DAFNEplus sessions, representing ‘moderate’ fidelity (see Report Supplementary Material 5, table 6). Fidelity was lowest for sessions: ‘dose adjustment for above-range blood glucose’ (27%), and ‘social support – who is there for you?’ (38%). BCTs were delivered, on average, in 56% (range: 0–100%) of intended instances (see Report Supplementary Material 5, table 7). Examples of low-fidelity BCTs include: information about antecedents (0%), behaviour substitution (0%) and non-specific incentive (0%). Objectively verified fidelity was on average also lower for DAFNE sessions (29%, range: 0–80%) (see Report Supplementary Material 5, table 6).

Regarding modifications to session delivery, in the qualitative interviews,32 most facilitators reported having benefited from facilitator training and scripts provided to enhance fidelity; however, many also reported tailoring the scripts and modifying content to fit their own style of delivery and to respond to participants’ needs, especially once they became more confident delivering novel course content. Qualitative analysis of free-text responses on the fidelity checklists identified that modifications to session content fell within three overarching themes (see Report Supplementary Material 5, table 8). First, 'amended/adapted activities', which included reported changes to: the examples and materials used, advice given, mode of delivery and sequencing of content and activities. Indeed, in interviews, facilitators also described needing to occasionally adapt the sequencing of content when questions arose relating to a topic that was planned to be covered later in the course, requiring them to deliver content from later sessions ahead of time.

Second modifications in the form of 'less emphasis on activities', for example, some facilitators reported spending less time on activities, not demonstrating activities, only covering activities with some of the participants, omitting certain aspects of the activity and/or deciding to omit the activity altogether. This was also echoed in facilitator interviews, where they reported occasionally spending less time on activities due to time pressures, to provide more time for participant discussion and to accommodate different speeds and styles of learning in their group.

Third, new additions, with new content, most often corresponding to the BCTs: practical social support (n = 17) and information about health consequences (n = 8). Note, all BCTs additionally delivered in a session, they were not originally intended to be delivered in, were included in the intended content of other sessions of DAFNEplus, so they were not entirely novel. Additional reasons facilitators provided in the checklists for modifying or not delivering content spanned all COM-B domains: psychological capability (e.g. forgetting); physical capability (participant fatigue and hypoglycaemia); reflective motivation (activity not seen as relevant or helpful by facilitators); automatic motivation (habit/reverting to DAFNE); social opportunity (lack of participant engagement); physical opportunity (lack of time/staff/equipment) (see Report Supplementary Material 5, figure 3, table 9).

1. What impact (practical and emotional) does the intervention delivery have on facilitators and their workloads; what resourcing and support would facilitators and their colleagues need to deliver DAFNEplus in routine clinical practice?

See Lawton et al.32 for full details.

In brief, while many facilitators shared initial reservations, all described how their experiences of DAFNEplus programme delivery had had a positive, transformative impact upon their perceptions and working practices. This transformation was enabled by initial training and supervision sessions; the confidence gained from using scripts to support novel programme content delivery; and experiences of delivering the programme and observing DAFNEplus principles being well received by, and having a positive impact on, attendees. Due to these positive experiences, facilitators described using some DAFNEplus techniques and curriculum content in their routine clinical care. This included incorporating the module on ‘monitoring long-term health’ into delivery of standard DAFNE courses and using action planning and positive/affirmative techniques in routine clinical consultations. While facilitators supported a national roll-out, they anticipated various attitudinal and logistical (e.g. workload) challenges. For further information about how these challenges might be addressed, see Lawton et al.32

1. Do any unintended consequences arise from the delivery and receipt of the DAFNEplus intervention, for participants and/or facilitators?
   • Interviews

As reported,31 participants described gaining a greater array of benefits from the course module ‘monitoring long-term health’ than had been anticipated when it was designed. The original purpose had been to motivate regular attendance at screening appointments. However, participants described an increased motivation to adopt and sustain the use of FIIT due to now understanding that, by doing so, their risk of developing diabetes complications would be extremely low. This important finding suggests that, even if the DAFNEplus programme is not rolled-out, serious consideration should be given to integrating this module into existing DAFNE courses.31

• Psychological process measures

Improvements in HbA1c in the DAFNEplus (but not DAFNE) group were also significantly associated with the worsening of participants’ perceptions of the convenience of their diabetes management (DME-Q subscale).

1. What are DAFNEplus participants’ experiences of, and views about, transitioning from receiving face-to-face consultations and support to remote/telephone support? What are the implications of this transition on how they think about and self-manage their diabetes and their perceived need for input and support from health professionals? What are their preferences for care delivery (virtual vs. face to face) moving forwards?

Interview participants recognised the importance of moving to remote care delivery when physical distancing measures were in place due to COVID-19 and were appreciative of being able to complete the programme despite the onset of the pandemic. Participants did not believe that the quality or content of remote IS sessions had been compromised because Glucollector and other technology [e.g. Freestyle Libre (Abbott Diabetes Care, Witney UK; 2020) and Microsoft Teams (Microsoft 2021)] permitted real-time joint review of glucose patterns. However, moving forward, most expressed a preference for reverting to face-to-face consultations, although they often struggled to articulate why. Some suggested that face-to-face consultations felt more personal and enjoyable, in large part because they felt they had forged strong relationships with facilitators during their face-to-face courses.

1. What impact has the COVID-19 pandemic had on how DAFNEplus participants think and feel about their diabetes (e.g. diabetes-related distress)?

Most participants who were in employment moved to home-based working or were furloughed when data collection took place for the interview study. Most indicated that this transition had had a positive impact on their diabetes self-management because they had more time to focus on themselves and their diabetes and they were more likely to lead routinised lives during lockdown periods. A minority, however, described high levels of distress resulting from physical distancing measures and/or knowing that, because they had T1D, they were at higher risk of developing serious/potentially life-threatening health problems in the event they got severe acute respiratory syndrome coronavirus 2. These participants described struggling to focus on their diabetes self-management due to high levels of anxiety and/or low mood.

Health economic evaluation

The economic evaluation had four components: pre-trial modelling of potential cost-effectiveness, the expected value of sample information (EVSI) of continuing the RCT after the COVID-19 pause, the main EEACT and long-term extrapolation of health economic outcomes using the Sheffield Type 1 Diabetes Policy Model.

Pre-trial and pre-trial restart modelling was conducted, and this is presented in Appendix 5. Appendix 6 details the cost-effectiveness of DAFNEplus compared to DAFNE in a population with T1D in a UK NHS setting using within-trial analysis methods.

Economic evaluation alongside the trial

Methods for economic evaluation alongside the clinical trial

A health economic analysis plan was developed prior to starting analysis and is available as a supporting document (see Report Supplementary Material 6). The primary analysis is the EEACT. A secondary analysis was also undertaken using long-term extrapolation. This is a switch around from the protocol because the trial found no evidence of a difference in biomedical outcomes (see Randomised controlled trial results, Key findings: biomedical outcomes).

Population

Details of the inclusion and exclusion criteria for the DAFNEplus trial are available in Workstream 4: cluster randomised controlled trial with concurrent process and health economic evaluations, Participants.29 The population for the health economic analyses was the full ITT population. The setting was in the UK NHS. The intervention was DAFNEplus. The control was DAFNE.

Perspective

Following NICE guidance,39 we took an NHS and personal social services’ perspective. The time horizon of the analysis was 1 year (study duration). No discounting of costs and QALYs were applied. We calculated costs, QALYs from the data. ICERs were calculated from the between-arm differences in costs and QALYs.

Quality-adjusted life-years

The QALYs were calculated from EuroQol-5 Dimensions, five-level version (EQ-5D-5L) responses from participants and in the trial and the Hernandez Alava et al.40 algorithm using the area under the curve method.

Healthcare resource use

We used standard national sources to calculate the unit costs of any reported resource use in the study. Costs were calculated in 2021–2 GBP, with costs from previous price years inflated to 2021–2 prices using the NHS cost inflation index (2015–6 onwards) and Hospital & Community Health Services pay and prices indices (before 2015–6).

Course costs

These were calculated using a survey of education courses at trial sites; records of structured follow-up contacts and expert opinion on technology costs if DAFNEplus were to be implemented. Healthcare resource use and medication use were collected in the study using self-reported questionnaires from study participants. Total costs within the trial period were calculated adding total medication costs, total healthcare resource use costs and total course costs.

Analysis

Data were cleaned and processed in R statistical software version 4.4.1 (The R Foundation for Statistical Computing, Vienna, Austria). To conduct imputation and fit the statistical models, we used Stata^® (StataCorp LP, College Station, TX, USA) version 18. Missing cost and QALY data were imputed by using multiple imputation using chained equations using a predictive mean matching procedure. Fifty-five imputation iterations were used, as approximately 55% (239/423) had missing cost, QALY or covariate data. Within this group, with any missing data, 213/423 people had missing QALY data [missing EuroQol-5 Dimensions (EQ-5D) at baseline, 6 months or 12 months], and 193/423 people had missing cost data. We fitted a seemingly unrelated regression model, with cluster-adjusted standard errors, to cost and QALY outcomes to calculate the effect of DAFNEplus compared to DAFNE. ICERs were calculated, which are the change in costs divided by the change in QALYs. These are then compared to the normal range for a cost-effectiveness threshold in the NICE methods guide of £20,000–30,000 per QALY gained. Uncertainty in incremental costs and incremental QALYs was undertaken using the coefficients and covariance matrix from the seemingly unrelated regression.

Results for economic evaluation alongside the clinical trial

The base-case result is given in Table 2. The incremental cost of DAFNEplus versus DAFNE was estimated at £611 per person randomised. The incremental QALYs gained were estimated at + 0.008. The ICER of DAFNEplus versus DAFNE in the base-case analysis is £78,147 per QALY gained, well above the typical NICE thresholds of £20,000–30,000 per QALY gained.

This result of being over £20,000 per QALY gained was robust to 11 different scenario analyses, which varied the utilities, costs, assumptions about missing data (e.g. complete-case analysis only) and population subgroups. Details are provided as in Appendix 6, Table 10.

The original Health Economics and Decision Modelling Analysis Plan is embedded in Report Supplementary Material 6 as a pdf object. This covers both the EEACT and the long-term model extrapolation using the Sheffield Type 1 Diabetes Policy model version 1.5.

Patient and public involvement

Aim

The overall aim with the PPI groups in DAFNEplus was to integrate members throughout the programme, seeking their lived knowledge, advice and review of the WS activities into the programme of research. PPI activity was funded from the outset of the grant.

Methods

Patient and public involvement members were engaged from Sheffield and London to two patient advisory groups (PAGs) as representatives of the wider T1D population. PPI members were identified via an online post through the DAFNE User Action Group (a national representative group of those who have completed a DAFNE course), the STH Diabetes and Endocrinology Patient Panel, which includes patients from a range of backgrounds, and advertising by the local clinical team.

Rather than offering formal research training for the PPI members involved in the study, which can be considered as a barrier to some patient groups, DAFNEplus training was developed and delivered by the PPI study lead, Wendy Baird, following consultation with individual members. We believed that it was important that service user groups were given time to establish ground rules and rapport both with each other and the researchers. It was hoped this would encourage greater participation, as service users increasingly became integral and valued members of the team. Expectations from the perspective of both the researchers and service users were clarified at the outset. In addition to a team experienced in PPI, a trials unit research assistant (DB) experienced in supporting patients and the public acted as a dedicated point of contact for service users.

Proposal development

Recommendations for further research in collaboration with the original DAFNE programme PPI group (15 members) informed development of the DAFNEplus proposal. Both the Sheffield PAG and the STH Endocrinology Patient Panel reviewed the proposal. Two DAFNE graduates were co-applicants on the grant.

Intervention development – workstreams 1–3

Author WB led the PPI involvement; DB facilitated involvement of all PPI members. Workstream leads were asked regarding their requirements from PPI.

The development and piloting stages of the programme were an intense period of evolving review and reflection. Regular structured meetings and review sessions with clear aims and objectives were held throughout the 2-year development and piloting stage. These processes involved review of the evolving DAFNEplus content, documentation and technical developments from specialist clinicians and academics from the universities of Sheffield, Edinburgh, Surrey and UCL. Processes were complex and iterative, involving cycles of development across workstreams.

The PAG members contributed to an interdisciplinary consensus exercise to inform the identification of behavioural content and principles for implementation in the developing DAFNEplus curriculum.

During the pilot studies, the PAGs worked in close collaboration with the CWG. Meetings were scheduled to facilitate the input of the PAG groups into the developing matrices of the CWG activity.

Cluster randomised controlled trial

The nature and intensity of the PPI activity during the RCT phase of the programme naturally decreased. Three meetings were held during this time, and four newsletters were published with the aim of updating the group regarding progress with the trial. Meetings were face to face where possible; during COVID-19 they had to be video calls, and some were hybrid. Two lay members (AD and LD) contributed to the Programme Steering Committee for the duration of the trial.

Patient and public involvement study results

Patient and public involvement input in the proposal development stages strongly advocated the need for researchers to understand patient perspectives. This ensured that issues which were a priority for patients were included in the DAFNEplus proposal.

Working with clinical and academic colleagues, the PAG groups identified the BCTs that were likely to drive sustained behaviour change via the structure of the course and the language used.

Substantial contributions were made to the development and piloting of the revised course curriculum, technology testing and development and participant materials such as information sheets.

During the trial proceedings, the PAG contributed to the trial management via the Trial Steering Committee. Members have been active in dissemination throughout the research programme, including blog posts, contribution to the lay summary and other aspects of the trial. Interestingly, when we discussed the results of the clinical trial with our PPI group, they were unsurprised that there were no significant differences between standard DAFNE and DAFNEplus in terms of HbA1c. They were more interested in the patient-reported outcome data, which showed greater reduction in diabetes distress and improved diabetes-specific QoL. There overall view was that, where practical, the DAFNE Collaborative should consider incorporating some of the DAFNEplus components of the curriculum into future DAFNE courses.

Patient and public involvement discussions/conclusions

Patient and public involvement knowledge and involvement throughout the programme ensured that the study took account of participants’ preferences, coproducing a tailored DAFNEplus system tangible for all patients. The linking of the CWG and PAG activity provided an innovative method of iterative and organised linked working.

Patient and public involvement reflections critical perspective

Our approach to PPI resourcing and planning ensured that the patient representative voice was key during the intervention development. It allowed the establishment of strong and supportive relationships with our PAG members, and develop key processes, including our PPI activity schedule and reporting log. This was important when faced with challenges such as the COVID-19 pandemic and changes within the academic team.

Consideration of equality, diversity and inclusion

Consideration of equality, diversity and inclusion was considered throughout the programme of research, which was initiated in 2016.

The proportion of different ethnicities in T1D diagnosed in childhood and young people (which includes the large majority of those with the condition) was recently reported in Fallon et al;41 84% were White, 6.7% were Asian, 3.9% were Black, 3.1% were mixed ethnicity and 2% were other. There was almost no effect of deprivation on the prevalence of T1D in contrast to type 2 diabetes where prevalence was around 10 times higher in those most deprived. In the DAFNEplus trial, the number of those reporting White ethnicity was over 85% in both groups, although it was higher in the DAFNEplus group (87% vs. 94%), probably due to centre differences across England and Scotland. There was a more even spread of educational qualification apart from those with no qualifications.

We did not collect detailed information on access to diabetes technology in relation to deprivation, an important limitation as it was particularly relevant to the DAFNEplus intervention. A recent paper has reported the impact of socioeconomic deprivation on access to technology in three urban hospitals in the UK.42 And, it was notable that the percentage of adults using real time CGM was significantly greater in the least deprived.41 Although no association between glucose levels (as measured by HbA1c) and deprivation was found, it is clear that greater efforts need to be made to address inclusion. As we discuss elsewhere, one of the key components of the DAFNEplus intervention was using technology to reflect on glucose patterns and then adjust the insulin dose and timing. Furthermore, both DAFNEplus and standard DAFNE include an extensive written handbook in English. Since reasonable proficiency in both written and spoken English was an inclusion criterion, this inevitably reduced access to the DAFNEplus trial from ethnic minorities.

We began discussing how we might improve the effectiveness of the DAFNE intervention following the completion of both our first programme grant and the REPOSE trial1 in 2017. Since then, the diabetes research community have become increasingly aware that the introduction of technologies, such as CGM, and more recently, hybrid closed loop, may accentuate the barriers to successful self-management in T1D among those from deprived communities. As a result of our experience in delivering this programme grant, in our ongoing work, we are now attempting to address these unmet needs among ethnic minorities and deprived communities. We have already connected with ethnic minority community groups and primary care networks delivering care in the most deprived areas of Sheffield and London. We are actively involving them in our current research work by ensuring we recruit significant numbers to contribute both to our PPIE groups and to our groups participating in our experienced-based co-design.

Discussion

Summary of findings

Over the 8 years of the programme, we completed Phase 1 (WSs 1–3 and a pilot feasibility trial) successfully and on time. In WS1, we applied behavioural science to develop the DAFNE structured education programme in creating the DAFNEplus curriculum. The work in WS2, provided a high-level logic model upon which to base follow-up requirements for the trial, and a model of intervention development, the FUSED model. In WS3, we developed (1) technologies to transmit data from individuals with T1D to support shared visualisation for patients and clinicians based on collected data from various monitoring devices and (2) interactive online modules specific to people with T1D that could support self-management by consolidating learning and promoting behaviour change.

Following a successful pilot and feasibility trial, we moved to Phase 2, a cRCT in 13 DAFNE centres, 7 of whom delivered the original DAFNE course and 6 who delivered DAFNEplus in which we anticipated a primary analysis population of 422 participants.

Recruitment was back on schedule when, in March 2020, we experienced a major disruption due to the COVID pandemic. All face-to-face sessions and recruitment were suspended for 18 months from March 2020 to end of September 2021 and when it restarted, recruitment thereafter was much reduced but was completed with reduced numbers in November 2022. We eventually recruited 471 participants, of whom 309 were included in the primary ITT analysis population (176 at DAFNE sites, and 133 at DAFNEplus sites). This was over 130 less than originally planned, but we still had adequate statistical power to test our primary end point.

Biomedical outcomes

In the primary ITT population (those with a HbA1c of > 7.5%), mean HbA1c reduced from 9.1% to 8.2% (75.9–66.5 mmol/mol) in the DAFNE group and from 9.1% to 8.1% (75.6–65.1 mmol/mol) in the DAFNEplus group. There was no evidence of a between-arm difference in HbA1c at 12 months {point estimate [95% CI −0.014 (−0.284 to 0.255); p = 0.915]} in the primary ITT population. No evidence of between-group difference was found for any of the secondary end points.

In the full ITT population (245 DAFNE, 178 DAFNEplus) for those where we have paired baseline and 12-month HbA1c data, around 68% improved their glucose levels during the trial; 46% DAFNE, 48% DAFNEplus achieved HbA1c < 7.5%; 45% DAFNE, 49% DAFNEplus lost ≥ 0.5%; and 24% DAFNE, 28% DAFNEplus achieved both measures at 12 months.

During follow-up, at least one episode of severe hypoglycaemia was experienced by 3% in the DAFNEplus arm (vs. 7% in the DAFNE arm) in the full as-randomised population.

Psychological outcomes

In the ITT analyses, the adjusted mean difference in the impact of diabetes on QoL [i.e. the ADDQoL-15 AWI score at 12 months (primary psychological outcome)] was 0.2 (95% CI −0.1 to 0.5; p = 0.21), equivalent to a standardised effect size of d= 0.1 (95% CI −0.1 to 0.4). Similarly, in the PP analysis, the adjusted mean difference of the ADDQoL-15 AWI score at 12 months was 0.4 (95% CI −0.01 to 0.9; p-value= 0.06), equivalent to a standardised effect size of d= 0.2 (95% CI −0.1 to 0.6)

For the secondary psychological outcomes, at 12 months, there was a statistically significant difference in diabetes distress (PAID: adjusted mean difference −3.3, p = 0.001 and d = −0.4) and hypoglycaemia avoidance (HFS-II behaviour subscale: adjusted mean difference −1.1, p = 0.002 and d = −0.3), and diabetes-specific positive well-being (4-item subscale of the WBQ-28: adjusted mean difference 0.9, p = 0.004 and d= 0.4), favouring the DAFNEplus intervention.

Our qualitative research demonstrated that DAFNEplus was well received by both participants and HCPs and that all components of the intervention were well liked. These findings also suggested that DAFNEplus worked as intended, which is consistent with the programme theory.38 However, many participants also highlighted a need for ongoing, tailored support from DAFNEplus trained HCPs to further consolidate and/or retain programme benefits over time

Health economic outcomes

Our primary base-case within-trial analysis using EEACT methods showed that DAFNEplus had an incremental cost of £611 versus DAFNE, an incremental QALY of + 0.008 and an ICER of £78,147 per QALY gained. All our scenario analyses, including complete-case analysis and conducting a long-term extrapolation model of biomedical outcomes had ICERs > £20,000 per QALY gained. NICE’s typical range for a maximum acceptable ICER is £20,000–30,000 per QALY gained, and the ICER for DAFNEplus versus DAFNE is highly likely to exceed this range.

Reflections on what was and was not successful in the programme

During Phase 1, we delivered WSs 1–3 and a pilot feasibility successfully. We demonstrated that we could use the data from our systematic review and using the logic model, used a FUSED approach to develop the three main components of our intervention, incorporating behaviour change science, technology to collect blood glucose data (obtained from both blood glucose monitoring) and CGM (obtained from flash glucose monitoring) and clinical psychological principles.

We showed no statistically significant between-group differences at 12 months in terms of HbA1c, severe hypoglycaemia or other biomedical outcomes in the impact of diabetes on QoL, and there were few adverse events. This suggests that DAFNEplus is equivalent to DAFNE in terms of its impact on both glycaemia and QoL. A large majority of participants (from both DAFNE and DAFNEplus groups) gained from participation in the DAFNEplus trial, between 60% and 70% achieving glucose levels closer to target. Although our primary psychological outcome, diabetes-specific QoL, improved to the same extent in both groups, this was observed both groups. What is unclear (and discussed in the following section) is to what extent the COVID pandemic may have affected the cRCT outcomes.

Importantly, the absence of a control arm in which structured training was not provided means that we cannot be certain that participation in training explains the decreases in HbA1c and severe hypoglycaemia in both groups. Yet, it seems unlikely that mere trial participation led to sustained decreases in HbA1c and severe hypoglycaemia for up to a year, particularly as allocation to such a control arm, in our original RCT involving the DAFNE intervention, showed no effect on HbA1c. Furthermore, real-world data which are collected as part of an annual audit, involving global DAFNE centres, currently show that around 50% now achieve a HbA1c of 48 mmol/mol (per com Elliott J. chair of DAFNE Executive, March 2025). This compares to only 1–2% of trial participants achieving this level of HbA1c in 2017, when the REPOSE RCT, comparing the effectiveness of DAFNE training in pump versus multiple injections was published.1 These data indicate that currently DAFNE training, whether delivered as standard DAFNE or DAFNEplus (probably allied to the greater availability of CGM technology), now enables adults to achieve glucose levels which are far closer to nationally recommended targets than was the case when the DAFNEplus grant application was being planned and written (see below for further discussion).

In Challenges and methodological limitations relating to the method or execution of the research, we have speculated why the participants in both arms of the RCT reached glucose levels which we had aimed to achieve in DAFNEplus alone. We also highlight and discuss the significant and clinically relevant psychological benefits which we observed in those undertaking the DAFNEplus arm of the trial and which we had not anticipated.

Overall, the fidelity of delivery of DAFNEplus was moderate, indicating that the intervention was mostly delivered as planned. In particular, DAFNEplus sessions, including more novel content/BCTs, were delivered with moderate-to-high fidelity, suggesting there was a reasonably fair test of the intervention in the cRCT. DAFNEplus was not 100% fully and consistently delivered as intended and thus internal validity of the cRCT is to an extent impacted. However, given the highly complex nature of DAFNEplus, requiring facilitators to deliver a large number of activities/BCTs in each session, fidelity was arguably realistic and acceptable. Nonetheless, the DAFNEplus fidelity levels are, on average, higher than those observed for other comparable complex interventions delivered in real-world healthcare practice settings.43 The activities not consistently delivered, and the facilitator modification/additions to content, highlight potential considerations for refining and optimising DAFNEplus ahead of any future roll out, scalability or replication.

The interview findings highlight the need for subgroup analysis. In our exploratory (latent profile) analysis44 (see Appendix 7, Figure 7), we used a data-driven approach and identified two clusters: the first cluster was more disadvantaged at baseline (in terms of higher HbA1c and greater negative impact of diabetes on their QoL), and they experienced steeper improvements over time than those in the second cluster, who experienced smaller improvements over time. Importantly, we saw that the first cluster (including more than half the sample) experienced statistically significant added benefit from DAFNEplus (compared to DAFNE) on their QoL at 12 months, but not on their HbA1c. As for the interview findings, these subgroup analyses show strong benefits from DAFNEplus for part of the sample, but not necessarily for everyone. This suggests that personalised approaches may be needed when discussing programme choice with adults with T1D – though it will be important not to make assumptions about who will benefit.45 We have not reported these subgroup findings in this report (as it was not pre-specified), but we intend to publish these in due course.

Challenges and methodological limitations relating to the method or execution of the research

1. Delay in starting recruitment of Phase 2, the cRCT. We originally planned to start recruitment on DAFNEplus at the start of September 2018, ending in December 2019. The introduction of the new SOECAT-driven processes for providing funding for excess treatment costs to NHS sites during this period created major delays for our cRCT as the process failed to anticipate the issues with cluster trials. The resultant discussions and negotiations with NHS England led to a delayed start at nine sites (the other four sites opening between October 2018 and January 2019). During this delay, glucose monitoring was increasing rapidly among people with T1D in the UK. By allowing both those with diabetes and professionals who support them to view and reflect on glucose patterns this technology was, in part, addressing the challenges which our technology WS (WS3) had been designed to overcome.
2. The effect of the COVID-19 pandemic. The other unexpected challenge was the COVID pandemic which had a major impact on both delivery of the intervention and follow-up during the cRCT. Having been on course to complete recruitment on our revised schedule, we were forced to suspend recruitment in March 2020 as both arms required face to face delivery to groups of six to eight. We were also unable to collect some of the primary outcome data as participants were could not attend clinics for collection of HbA1c samples which were designed to be measured in a central lab. We tried to mitigate this problem during lockdown by obtaining kits enabling participants to collect capillary blood samples at home and posted to laboratory to enable analysis. But, this inevitably reduced the number of measurements at both 6 and 12 months. The withdrawal of one site led to a slight decrease in the power of this trial, though this remained at just over 90%. We obtained a further costed extension for 18 months to enable us to complete and analyse the cRCT. Recruitment finally was completed in October 2022 and 12-month follow-up in October 2023 (around 4 years in total).

We have speculated on the effect of these delays and the pandemic on the results of the trial.

1. The use of flash and CGM. At baseline, 73 (41%) of those in the DAFNE group and 67 (51%) of those in the DAFNEplus group were using this technology. At 12 months, the numbers were 105 (80%) and 67 (68%), respectively. Since study participants had the opportunity to initiate diabetes-related technology irrespective of the trial (e.g. Freestyle Libre), this may have diminished the potential benefit of DAFNEplus on the study outcomes. Proportions would have likely been much smaller had we been able keep to our original timetable.
2. The introduction of telehealth due to the pandemic may have led DAFNE centres to be more responsive and flexible in their follow-up support (partly, the intention behind the DAFNEplus). As in (1), this may also have reduced the relative benefit of DAFNEplus.
3. The effect of the pandemic on the participants. Lockdown and isolation may have led to less focus on the self-management of diabetes in both groups or alternatively may have resulted in more focus on effective diabetes self-management.
4. The challenges for participants in providing samples for measurement of HbA1c (despite attempts to mitigate by arranging measurement at local sites where possible or providing kits to enable collection of samples at home) led to some missing data for our primary end point.
5. Missing data for economic analyses. Approximately, 55% of people had some missing data in the economic analysis in cost, QALY or control covariates (e.g. CGM status). Despite imputing the missing data using as much self-report data as possible to impute the missing data (e.g. if someone was missing their 6 months EQ-5D, their baseline and 12-month EQ-5D utility scores were used in the imputation procedure), this is still a relatively high amount of missing data, which could introduce some bias into our results. However, given our consistently high ICERs across all scenarios, the judgement of the health economics team is that it is unlikely that there is sufficient bias to decrease the estimated to ICER to under £20,000 per QALY gained.

Conclusions from the whole programme

Since there were no differences in HbA1c between DAFNE and DAFNEplus at baseline and 12 months, our hypothesis that DAFNEplus would lower HbA1c to a greater extent than standard DAFNE was unproven. HbA1c fell in both groups by between 0.9% and 1%, a fall which is around twice that seen in previous RCTs and what we had anticipated we would show in DAFNEplus alone. Thus, what was both surprising and perhaps, of most interest, in relation to the biomedical outcomes was the significantly greater success of those undertaking standard DAFNE courses in reaching target glucose levels than previously observed in our previous research programmes.

Whether the increasing use of CGM in both groups which allows people with diabetes to view their glucose levels in real time did result in participants reflecting more actively on their glucose readings and taking action is unclear. But, one of the key components of DAFNEplus was using technology (through Glucollector) to facilitate the ability of those undertaking self-management training to access and reflect on their glucose patterns. We have, therefore, speculated that the delays in both starting the overall programme of work and the 18-month break in the trial due to the COVID pandemic, leading to a much greater uptake of CGM in both trial groups was probably the main reason for comparable and larger improvement in HbA1c than we anticipated. We collected no data in the standard DAFNE centres to inform us to what extent they undertook remote consultations during the pandemic. Thus, the question as to whether remote support (albeit unstructured) also contributed to comparable HbA1c values during the trial remains unclear. We have also debated whether we perhaps should have focused on identifying fewer but important self-management behaviours in developing the DAFNEplus intervention. Successful self-management of T1D involves a huge number of relevant behaviours and perhaps we should have spent more time with both professionals and participants, agreeing those behaviours that they considered key to keeping glucose levels close to target.

Psychological findings

Another unexpected outcome of the trial was in relation to the psychological outcomes, both qualitative and quantitative. Positive outcomes of the trial and of DAFNEplus overall are therefore focused on the psychological outcomes.

There were no differences in overall QoL between DAFNE and DAFNEplus participants. Secondary psychological outcomes, which favoured DAFNEplus included: diabetes-specific distress and diabetes-specific positive well-being (PAID and W-BQ28 subscale) and hypoglycaemia avoidance behaviours (HFS-II). These suggest that the DAFNEplus offers benefits to participants in terms of reducing diabetes-specific distress and increasing the sense of satisfaction derived from managing diabetes as well as reducing the hypoglycaemia avoidant behaviours (without any increase in severe hypoglycaemia).

Within-group analyses showed significant and strong improvements in most psychological processes among those in the DAFNEplus group from baseline to course completion (largely maintained over 12 months), which were consistent with the proposed logic models.38 These included improvements in enhancing self-regulatory skills or behaviours (ability to develop, implement and flexibly maintain planned behaviour), enhancing diabetes strengths and resilience (behaviours and attitudes related to living with the demands of T1D), enhancing beliefs about own capabilities in self-managing T1D (confidence in diabetes), satisfaction with diabetes management, including effectiveness and intrusiveness. These findings align with the interview findings and suggest the intervention in that regard has been successful. However, the change in HbA1c over 12 months was associated with less convenience of diabetes management (DME-Q subscale), suggesting that noticeable efforts are still required to improve glycaemia, which are not resolved by structured education, technology or individual follow-up. These findings, captured in the quantitative psychological measures, are supported by the interview study findings.

However, it is important to point out that the improvements in psychological outcomes and processes seen in the DAFNEplus group were also observed in the DAFNE group. Further, in some cases, the DAFNE group had significantly stronger improvements over time compared to DAFNEplus. Some important methodological considerations may explain these findings, as described below. For example:

• There were substantial missing data in the quantitative data, potentially biasing the observed findings.
• Some constructs of importance to the logic model38 were not, or only partially, assessed, such as social support, motivation and personality traits.

Conclusions from qualitative studies

The interviews identified three groups of participants: pessimists, perfectionists and optimists. The first two groups reported multiple positive changes from the novel components of the DAFNEplus course, while the third group noted limited/no personal gains. This is important for several reasons:

• Firstly, personality traits where not assessed among the psychological measures, but they may have important predictive value. If, for example, x% of participants had no added benefit from DAFNEplus, it is possible that their responses would contribute to averaging out potential true group differences. Personality measures are inherently lengthy and would have added substantial participant burden. Furthermore, it was considered that asking people with T1D to complete personality assessments might contribute to diabetes-specific distress and stigma (e.g. feeling like any suboptimal outcomes are being blamed upon their personality type).
• Secondly, the interview findings strongly highlight the need for subgroup analysis. In our exploratory (latent profile) analysis which was not pre-specified (see Appendix 7), we used a data-driven approach and identified two clusters: the first cluster was more disadvantaged at baseline (in terms of higher HbA1c and greater negative impact of diabetes on their QoL), and they experienced steeper improvements over time than those in the second cluster, who experienced smaller improvements over time. Importantly, we saw that the first cluster (including more than half the sample) experienced statistically significant added benefit from DAFNEplus (compared to DAFNE) on their QoL at 12 months, but not on their HbA1c. As for the interview findings, these subgroup analyses show strong benefits from DAFNEplus for part of the sample, but not necessarily for everyone. This suggests that personalised approaches may be needed when discussing programme choice with adults with T1D – though it will be important not to make assumptions about who will benefit. We have not reported these subgroup findings in this report (as it was not pre-specified) but intend to publish these in due course.

Recommendations for future research

1. Although the findings of this research may in part, relate to our already described study limitations, they may require further evaluation in future studies, in particular benefits in reducing diabetes distress. This will be important in helping decide whether and if so, which areas of DAFNEplus, should be included in future courses.
2. Use of DAFNE training to support the roll-out of automated insulin delivery, particularly for those from deprived communities and ethnic minorities who have, to date, generally not participated in the RCTs and observational studies reporting benefit.
3. Whether including individualised structured support post DAFNE can reduce the number of participants in courses who struggle and so actually reduce the workload of hard-pressed nurses and dietitians. Nurse/dietitian educators currently spend many hours of time, seeing and supporting individuals who find the challenges of self-management beyond them and who are repeatedly referred back to diabetes educators by doctors. This will require a specifically designed study, measuring both feasibility and cost-effectiveness.
4. Published benefits of the DAFNE intervention and confirmation in this research programme that around 70% of those participating in a DAFNE training course (whether DAFNE or DAFNEplus) experience clinically relevant improvements in their glucose targets and major falls in episodes of hypoglycaemia or DKA; many centres across the UK fail to offer or encourage adults with T1D to participate in self-management training. Further research could explore why diabetes professionals fail to encourage participation in an intervention which is cheap, cost-effective and provides considerable improvement in both clinical targets and QoL.

Implications for practice and any lessons learnt

The DAFNE Collaborative, originally consisting of three participating centres during the first RCT in 1998, now includes over 60% of centres which support delivery of care to adults with T1D across the UK. It has an active research programme, including the original DAFNE RCT, the REPOSE trial1 and data collected, which include an annual national audit, real-world data collected by the DAFNE coordinating centre, annually. When our group assembled our research team to design and write the DAFNEplus Programme Grants for Applied Research (PGfAR) application in 2013 (12 years ago), ongoing research (including our own) was still demonstrating that most adults with T1D were running HbA1c levels far from target (confirmed by the REPOSE trial). Twelve years later, the situation has changed, certainly for DAFNE graduates.

As we discussed above, the DAFNE Collaborative now conducts an annual audit of those delivering DAFNE courses in the UK, reporting that 48% of DAFNE graduates who participate in standard DAFNE courses across the country reach HbA1c levels < 58 mmol/mol (7.5%). These results, reflected in the DAFNEplus RCT indicate that the DAFNE structured training course, now result in falls in HbA1c which are around double of that observed in previous RCTs and comparable to those in the DAFNEplus trial.

Since the Collaborative co-ordinates the delivery of training across the UK, we could implement developments in the DAFNE programme at speed in over 60% of centres across the country. Our health economic analysis shows clearly that the ICER in the base-case analysis is £78,147 per QALY gained, well above typical NICE thresholds, and show that based on comparable falls in HbA1c and severe hypoglycaemia and EQ-5D data, implementing DAFNEplus will not be cost-effective.

Our qualitative research demonstrated that DAFNEplus was well received by both participants and HCPs and that all components of the intervention were well liked. These findings also suggested that DAFNEplus worked as intended, which is consistent with the programme theory.38 Furthermore, many participants also highlighted a need for ongoing, tailored support from DAFNEplus-trained HCPs to further consolidate and/or retain programme benefits over time.

It is noteworthy that the fall in HbA1c observed in both groups [9.1–8.2% (76–66 mmol/mol)] is considered as an important and clinically relevant fall in reducing the future risk of microvascular complications. The spread of CGM now enables adults with T1D to reflect on their glucose profiles and implement self-management skills taught in both DAFNE and DAFNEplus more effectively than in the past. Although CGM and the increasing availability of semi-automated insulin delivery (delivered by hybrid closed loop insulin pumps) is transformational for many, our clinical experience, confirmed by this programme of work indicates that the demands of self-management in T1D remain considerable. Many diabetes professionals believe that these technologies which are increasingly available, and which will become more effective when combined with Artificial Intelligence (AI), will enable the vast majority of people with T1D to achieve target glucose levels.

The view of the DAFNE research group, which includes both health professionals and lay people with lived experience of T1D, is less optimistic. Over 100 years after the discovery of insulin as a life-saving treatment, the major limitations of insulin in therapeutic delivery remain apparent. Even the most modern insulin rapid-acting insulin analogues fail to lower glucose effectively for 20–30 minutes following sc injection, and insulin levels then remain raised for around 3–4 hours even if glucose falls below normal. This is highly unphysiological and explains why achieving near-normal glucose levels is so challenging. Although the addition of AI will doubtless improve insulin pump algorithms, it seems unlikely that sc insulin delivery will prevent excessive rises in glucose and the attendant risk of hypoglycaemia without a revolution in insulin development.

Arguably, we should also have focused more on the potential of DAFNEplus in improving psychological outcomes. At 12 months, there was a statistically significant and clinically meaningful difference in diabetes distress (PAID scale), hypoglycaemia avoidance (HFS-II behaviour subscale) and diabetes-specific positive well-being (4-item subscale of the WBQ-28), favouring the DAFNEplus intervention. We acknowledge that our primary psychological outcome, QoL, was no different in those randomised to DAFNE or DAFNEplus. Interestingly, in preparing this report, we have reread our original PGfAR application and were reminded that, in our original grant application, we originally chose diabetes distress as our primary psychological outcome. This was due to its increasing importance in reflecting the negative psychological impact of the considerable demands of self-management. However, we were required to replace it with a QoL, patient-reported outcome measure, which was suggested by some reviewers of our application and then requested by the PGfAR, Stage 2 panel.

Clearly, the COVID pandemic severely disrupted both the delivery and the analysis of the trial with 30% loss to follow-up for the primary biomedical outcome and 40% of the psychological end point. However, on balance, it seems most likely that the resulting delays and the rapid increase in technology probably account for the larger and comparable fall in HbA1c in both groups (also observed in the DAFNE audit). We have also reflected that in focusing on identifying the many behaviours required in self-managing T1D, we may have overestimated the likely benefit of including these in a redesigned curriculum without a greater understanding of those behaviours which are key. We have argued elsewhere that in requiring individuals with T1D to constantly implement a wide range of complex SCB, self-managing T1D is one of the most challenging conditions to manage. In reflecting on our results, perhaps we should have focused on a fewer number of key SCB and involved both those with diabetes and health professionals in more detailed discussions. We would advocate a more extensive debate with both those with lived experience and diabetes professionals when developing interventions to improve self-management where complex SCB are required.

In closing, we have identified those components that we are intending to implement within the national DAFNE programme going forward.

As we discuss above, diabetes professionals (nurses and dietitians) currently spend considerable time in repeatedly seeing adults with T1D who struggle to reach glucose targets, and arguably, much of it wasted. Implementing structured support post-DAFNE may benefit them (and the diabetes service) in using their time both more efficiently and rewardingly. This development (appropriately evaluated) and the clinically important reduction in diabetes distress are the prime areas for us to consider as we move the DAFNE training programme towards a more DAFNEplus format in future courses.

Finally, in the qualitative interviews, participants described increased motivation to adopt and sustain the use of FIIT due to now understanding that, by doing so, their risk of developing diabetes complications would be extremely low. This important finding suggests that, in addition to the components above, we should also integrate this module into existing DAFNE courses.9