A cluster randomised trial, cost-effectiveness analysis and psychosocial evaluation of insulin pump therapy compared with multiple injections during flexible intensive insulin therapy for type 1 diabetes: the REPOSE Trial

Type 1 diabetes mellitus and its treatment

People with type 1 diabetes mellitus (T1DM), around 250,000 individuals in the UK, have lost the ability to make insulin because of autoimmune destruction of the insulin-secreting β cells within the islets of the pancreas. Insulin is essential in the short term to prevent the onset of ketoacidosis, a potentially fatal condition. In the long term, the aim of insulin therapy is to keep blood glucose close to normal and so prevent the development of microvascular complications, such as retinopathy, neuropathy and diabetic kidney disease. Insulin is generally administered by intermittent subcutaneous injection, with the dose adjusted according to eating and other activities, such as exercise. Traditionally, insulin was given twice a day, often as premixed insulin, but such an approach imposes a rigid lifestyle and makes it difficult to maintain a glucose level close to normal. The need for intensification of therapy and its integration into flexible lifestyles is promoted in DAFNE (Dose Adjustment for Normal Eating) and other structured education courses. It involves giving quick-acting insulin just before eating and administering longer-acting background insulin, preferably twice daily, to maintain blood glucose levels in between meals. 1,2 This multiple daily injection (MDI) regimen involves a total of five or six injections a day. Blood glucose levels are monitored from finger-prick samples using a portable meter, and insulin dose calculations are based on self-assessed carbohydrate estimations on a meal-by-meal basis.

Insulin given subcutaneously cannot reproduce the physiological insulin profiles of non-diabetic individuals because of the limitations of insulin formulations and the site of delivery. The relatively slow rate of insulin absorption leads initially to postprandial hyperglycaemia, followed, 1 or 2 hours later, by an increased risk of postabsorptive hypoglycaemia, particularly during the night. Thus, keeping blood glucose close to normal can delay or prevent complications, but brings with it frequent periods of hypoglycaemia. These are categorised as mild, moderate or severe episodes, ranging from mild symptoms, self-managed by ingesting rapid-acting carbohydrate, through to greater disruption in daily routine due to cerebral dysfunction, through to major episodes of coma and seizure requiring third-party assistance. The inability of intermittent injection therapy to control blood glucose tightly without an attendant risk of hypoglycaemia results in many individuals keeping their blood glucose at higher than desirable levels. This leads to an increased risk of serious diabetic complications, which can affect the eyes, feet and kidneys. These complications, plus the associated high risk of cardiovascular disease, reduce both the length and quality of the individuals’ lives.

Insulin analogues

Short- and long-acting insulin analogues have slightly more physiological profiles than insulins of human or animal structure, but cannot reproduce those observed in people without diabetes. 2 Systematic reviews of clinical trials of insulin analogues involving people with T1DM have reported only minor advantages compared with human insulin, with a reduced risk of symptomatic hypoglycaemia, particularly at night. 3,4 This may be, in part, because those people who are at the greatest risk of hypoglycaemia are frequently excluded from clinical trials. Interestingly, in a recent crossover trial comparing MDI of human insulin with analogue insulin, the investigators specifically recruited individuals who had experienced problems with hypoglycaemia, and found that those using analogue insulin had significantly lower risks of severe hypoglycaemia, particularly at night. 5

Insulin pumps

There is clearly an urgent need for better methods of insulin delivery. Insulin pumps were first used clinically in the early 1980s, but randomised controlled trials (RCTs) conducted in the UK failed to show any clinical benefit. At the time, the technology was poorly developed, but has advanced considerably, particularly in the last few years. Insulin pumps are now the size of a small mobile phone and deliver insulin continuously under the skin via a small plastic tube and cannula [continuous subcutaneous insulin infusion (CSII)]. 6,7 These devices are filled with reservoirs of quick-acting insulin only (usually an insulin analogue), which provides insulin replacement by delivering both the mealtime and background insulin. When infused continuously at low rates they ‘mimic’ basal insulin secretion, and this is generally delivered more consistently and accurately than is achievable by the longer-acting insulins, particularly at night. The insulin boluses used to cover meals and correct high blood glucose levels are delivered much more rapidly. All of the insulin doses can be controlled by the patient, based on calculations similar to those required for insulin dosing with a MDI regimen.

The purchase and use of pumps is more expensive than MDI, with pumps at current prices costing around £2500 each, plus £1500 per year extra for running costs. 8 The marginal cost per annum over MDI is about £1800. 9 The potential advantages are more stable blood glucose levels, a reduced risk of hypoglycaemia and a more flexible lifestyle. Pump treatment may deliver insulin more effectively than MDI but does not provide a technological ‘cure’. The same competencies needed for successful insulin self-management, previously described for MDI, are required for pumps, but with additional skills required to operate the pump device itself. Thus, pumps are probably more useful to those individuals who are actively and effectively self-managing their diabetes rather than those who expect the pump to ‘manage’ their diabetes for them.

Pumps are currently used by around 40% of people with T1DM in the USA and > 15% in Europe. 10 In contrast, the proportion in the UK was around 6% in adults in 2012. 11,12 Proponents of pump treatment have proposed that far more patients should be offered treatment in the UK and that current policies are depriving many of the opportunity to improve glycaemic control, reduce hypoglycaemia and improve quality of life (QoL). 12 The UK’s National Institute for Health and Care Excellence (NICE) has recently extended recommendations for the use of pumps for adults with T1DM. The guidance suggests that pump treatment be considered for individuals who are experiencing problems with hypoglycaemia, particularly when this limits the ability to improve glycaemic control. NICE has noted the paucity of evidence for efficacy from RCTs. 13

Problems with evidence in National Institute for Health and Care Excellence appraisals

There have been two appraisals9,14 of pumps by NICE, both supported by technology assessment reports undertaken by some of the present authors, which reviewed the evidence on clinical effectiveness and cost-effectiveness. The first report14 noted that there were no trials of pumps against ‘best MDI’ with long- and short-acting analogue insulins; some trials had unequal amounts of education in the arms (with more in the pump arms); and the trials had focused on easily measurable outcomes such as glycated haemoglobin (HbA_1c), rather than on benefits in terms of flexibility of lifestyle and QoL. The report recommended trials of pumps against analogue-based MDI.

The second report9 found that few such trials had been done: one in children, not relevant to this work, and three in adults. Furthermore, the three adult studies15–17 presented data for a small number of participants who were followed over a short period only. The first of these studies was a 24-week pilot study15 in adults with altered hypoglycaemia awareness and debilitating hypoglycaemia. The three study arms consisted of seven patients each and compared (1) analogue MDI, (2) pump and (3) education and relaxation of glycaemic targets. All of the subjects were naive to analogue insulin use and some had never tried MDI, and so were not representative of the type of patients for whom NICE recommends pumps.

The second trial16 recruited 39 adults with T1DM, who had already been on pump therapy for at least 6 months, and who were randomised to stay on pump or switch to glargine (Lantus, Sanofi-Aventis, Guildford, UK)-based MDI for 4 months. The primary end point was glucose variability, which was 5–12% less with the pump. Despite this, there was no significant difference in the frequency of hypoglycaemic episodes or HbA_1c.

The third study17 studied 50 patients with T1DM from Italy, UK (Newcastle, Bournemouth) and France, who were naive to pumps and glargine, to which they were switched for the trial, having been previously managed on neutral protamine Hagedorn (NPH)-based regimens. Follow-up was for 24 weeks. Patients were randomised to pump or analogue MDI in an equivalence study. The difference in HbA_1c at the study end was only 0.1% (approximately 1 mmol/mol) and the costs with the pump were three times higher.

Thus, the evidence base from trials for comparing pumps and ‘best MDI’ remains weak in terms of numbers, with a total of only 103 patients and short-term follow-up. Furthermore, the patients in the trials were dissimilar to those considered suitable for a pump by NICE, which expects patients to have tried analogue-based MDI before using the pump.

Given the paucity of RCTs, the assessment group also looked at observational studies of adults in which a pump was clinically indicated, mostly because of the limitations of intermittent injections. This comparison has the advantage of measuring change in glycaemic control and hypoglycaemia in those who have most to gain, and these studies showed improved HbA_1c of the order of around 0.5% (5.5 mmol/mol). Interpretation of data from observational studies face limitations from bias, and, furthermore, of the 48 observational studies, only nine reported QoL. Study numbers were small and duration was usually short. The longest study noted that initial benefits from pumps might not be sustained.

Therefore, again, NICE was faced with an evidence base with considerable shortcomings, too few trials, durations too short, numbers too small and a need to use observational studies. A recent meta-analysis by Monami et al. 18 concluded that ‘available data justify the use of CSII for basal-bolus insulin therapy in type 1 diabetic patients unsatisfactorily controlled with MDI’. However, most of the RCTs in their analysis were NPH-based and the Bolli et al. 17 trial, with its negative result, was missed.

A systematic review of the cost-effectiveness of insulin pump therapy in adults with T1DM was conducted by Roze et al. 19 They identified four cost-effectiveness studies in the UK setting, three of which presented an incremental cost-effectiveness ratio (ICER). 9,14,20,21 The ICERs in these studies were £11,461 per quality-adjusted life-year (QALY) gained, £25,648 per QALY gained and £37,712 per QALY gained. Two out of the three studies had ICERs that lie within, or below, the £20,000 to £30,000-per-QALY-gained range that NICE usually uses to determine if a health technology is cost-effective. 22 These two studies did receive commercial sponsorship, whereas the study with an ICER of £37,712 was commissioned on behalf of NICE.

Rationale for the trial

We hypothesised that much of the benefit of pumps may come from the retraining and education in intensive insulin management, which allows patients to use pumps safely. 23 In many DAFNE centres, reimbursement for pump use is conditional on patients having attended a DAFNE education course and so some patients undertake DAFNE training with the intention of moving to pump treatment thereafter. It has been our clinical experience that many individuals decide not to switch to the pump after attending a DAFNE course, as they then realise that what they required was training in insulin self-adjustment rather than a different technical way of delivering insulin. Ray et al. 24 found that 69% of those being considered for insulin pump therapy stay on MDI after completing DAFNE. Importantly, trials and observational studies of high-quality training alone (with standard insulin injections) show benefits in blood glucose control, hypoglycaemia and QoL, which are as good, if not better, than those reported after pump therapy. 2,25,26

To our knowledge, no trials in adults, comparing pump treatment with modern MDI, used the same structured training in insulin adjustment, resulting in the added benefit of the pump technology remaining unclear. 23 There was an urgent need to establish this, and identify patients who benefit the most. A RCT was needed to establish these outcomes without bias.

The DAFNE course is a 1-week structured education course, teaching adults with T1DM the skills in insulin self-adjustment and carbohydrate counting. 2 DAFNE courses are currently delivered in more than 70 centres across the UK, with over 37,000 individuals (DAFNE graduates) now trained. We therefore set out to conduct a novel study in which adults waiting for a DAFNE course were randomly allocated to undertake either the standard MDI course or DAFNE incorporating use of pump therapy.

The investigators involved in this work have been undertaking research into other aspects of DAFNE for many years. During recent work funded by a National Institute for Health Research (NIHR) programme grant [Programme Grants for Applied Research (PGfAR)] we measured cost-effectiveness and identified which components of the course are crucial, as well as identifying the factors determining which DAFNE patients managed their diabetes more effectively. 27 This work included funding to pilot a combined DAFNE and pump course, which enabled us to develop a pump curriculum and associated pump-specific resources, ensure that the outcome measures that we wanted to use were feasible and estimate the likely recruitment and retention rates.

We then assembled a study group with expertise in structured T1DM education, pump therapy (having trained in total over 700 pump patients) and health economic assessment of diabetes interventions.

Decision problem: aim of the REPOSE Trial

The aim of our trial was to establish for patients, professionals and those funding the service, the added benefit of using a pump during intensive insulin therapy. We conducted a RCT comparing optimised MDI therapy (using rapid and twice-daily, long-acting insulin analogues) with pump therapy in adults with T1DM, for which both were provided with high-quality structured education (DAFNE).

Pump therapy

Table 1 shows the trials of pump therapy against MDI in adults with T1DM, excluding those in pregnancy. There have been only four trials of pump versus MDI with analogue insulin use in both arms, and the longest follow-up period was 24 weeks, which, as we will show in Chapter 3, is insufficient to achieve the full potential of pump therapy.

Table 1 shows that only five trials (assuming that Lepore et al. 44 is a trial – the paper does not mention randomisation but Misso et al. 31 in the Cochrane review say it was a RCT and that it had access to unpublished data) had a duration of ≥ 12 months, and none used analogue MDI. Lepore et al. 44 report HbA_1c only at baseline and 12 months. 44 Dahl-Jørgensen et al. 54 reported a steep drop in HbA_1c with a plateau after 3 months, but this reduction started in the 2-month run-in period before pump therapy was started.

Four trials15–17,40 used analogue insulin in both arms. The Hirsch et al. 40 trial had patients on pump and MDI for only 4 weeks. The Thomas et al. 15 trial was a pilot, with only seven patients per arm.

One new trial has been published since the last appraisal by NICE: Bruttomesso et al. 16 This trial recruited 42 patients already well controlled on the pump (mean HbA_1c 7.4% at randomisation) and randomised them to continuing pump therapy, or to MDI with lispro and glargine. The aim was to see if the need for pump therapy was reduced by the arrival of the analogue insulins. Patients had only 4 months on MDI. Three patients withdrew shortly after starting MDI because of poorer glycaemic control. After 4 months the patients switched to the other treatment arm. The primary outcome was glucose variability, as assessed by SMBG. HbA_1c during the study was 7.3% in both arms. There was no difference in the frequency of severe hypoglycaemia, but moderate hypoglycaemia was about 23% less frequent on pump therapy, although the definition of this is not stated in the published study. Glucose variability was 5–12% less on pump therapy, depending on time of day and method used. At the study end, patients could choose between pump therapy and glargine-based MDI. Thirty patients chose pump, five chose MDI and four opted for summer MDI and winter pump. The study was supported by Disetronic and one author worked for the company. 16

The Bolli et al. 17 trial (see Table 1) was published in 2009 but had been available in abstract form for the last assessment report.

Overall, therefore, there was still a poor evidence base with only one new trial, and that being of short duration (4 months on each arm) and limited sample size (only 39 patients).

Observational studies

Bacon et al. 55 reported 10-year follow-up data on 197 patients on pump therapy. The main indications for the pump were recurrent hypoglycaemia and poor control. HbA_1c improved by about 0.7% and the number of severe hypoglycaemic episodes by about 80%. Only about 5% discontinued pump therapy.

Beato-Vibora et al. ,56 from King’s College Hospital, looked back over 12 years of pump therapy in 327 patients, with a mean duration of 4.3 years on the pump. An initial reduction in HbA_1c of 8 mmol/mol or 0.7% was partially maintained with reduction at year 5 of 0.4%. The proportion of people having frequent mild-to-moderate hypoglycaemia fell from 29% to 12% and the frequency of severe hypoglycaemia was halved.

Bruttomesso et al. ,57 from the Veneto region of Italy, provide a retrospective study of all patients in their region who started pump therapy. Of 138 patients, 20 stopped pump therapy, although mostly in the earlier years. Strict eligibility criteria had to be met, including ‘the technical, physical and intellectual abilities’, plus motivation, stable personality and realistic expectations of pump therapy. All were familiar with MDI and received extra education. HbA_1c was 9.3% when starting pump therapy, fell to 7.9 by end of year 1 and was largely sustained there for 7 years.

Carlsson et al.,58 from Sweden, reported results of 272 patients with at least 5.5 years of follow-up. They compared their results with a much larger group on MDI. HbA_1c was reduced by 0.42% at 1 year and 0.43% at 2 years, but some of the effect was lost by 5 years when the reduction compared with the MDI group was only 0.2%. 58 A later paper59 reported that the reduction in HbA_1c varied by baseline levels, with a small reduction of 0.29% (85% CI 0.11% to 0.47%) in those with baseline HbA_1c of 7%, a reduction of 0.39% (85% CI 0.27% to 0.52%) in those with baseline HbA_1c of 8% and a larger reduction of 0.50% (85% CI 0.36% to 0.67%) in those with baseline HbA_1c of 9%, which would take them nowhere near target.

Cohen et al. 60 compared two cohorts from Melbourne in a non-randomised comparison. One group received pump therapy and the other received intensified MDI. Both were previously on analogue MDI. Among 126 patients on the pump, HbA_1c fell by 0.64% at 6 months, but then rose again, with a reduction of about 0.4% at 2 years and about 0.2% at 5 years. The reduction in HbA_1c on intensified MDI was smaller: 0.15% at 6 months. This was despite a similar programme of education, based on DAFNE but shorter, in both pump and MDI groups.

Lepore et al.,61 from three Italian centres, compared results in two matched groups of 110 patients on pump therapy and 110 on MDI, followed for 3 years. HbA_1c fell by 0.7% in the pump group and this reduction persisted for the 3 years. HbA_1c fell by 0.3% in the MDI group at 3 years.

Nixon et al. 62 reported a study of 35 patients on pump therapy. There was an initial fall of 1.7% in HbA_1c but by 5 years the reduction was only 0.9%. However, this reflected a mix of results, with one-third of patients reducing HbA_1c by 2.2% and maintaining it there, whereas others had no change on the pump or had an initial reduction not sustained.

Orr et al. ,63 from Ontario, report results among 235 patients on pump therapy. The overall baseline HbA_1c was 8.7%, which was reduced to 7.5% after 6 months on the pump, after which it drifted up again to 8.2% in years 3–10. In 39 patients who were followed for 10–15 years, the mean HbA_1c was 8.03%. However, two groups of patients who started with high baselines (8.5–10% and > 10%) reduced their HbA_1c to about 8% by 8–10 years.

Quiros et al. ,64 from Barcelona, followed 151 patients on pump therapy for 5 years. Overall, HbA_1c was reduced from a mean of 8.0% at baseline to 7.8% at 5 years. However, in the 61% of patients who started pump therapy because of poor glycaemic control, HbA_1c fell from 8.4% at baseline to 8.0% at 5 years. There was a marked reduction in severe hypoglycaemia. 64

Rosenlund et al. 65 from Denmark looked at the effects of 4 years of pump therapy on albuminuria compared with an unmatched group on MDI. On pump therapy, HbA_1c fell from 8.4% to 7.8%, maintained to 4 years. 65

Steineck et al. 66 from Sweden reported mortality data from a cohort of 18,168 people with T1DM in Sweden, of whom 13% were on pump therapy. Total mortality at 7 years was 6% in the pump group and 8% in the MDI group. There were many small differences that would increase the risk in the MDI group – more hypertension, more on lipid-lowering drugs, more with low physical activity, more smokers and more with low education levels. Steineck et al. 66 used propensity matching to adjust for the differences, and concluded that those on pump therapy had a 0.73 hazard ratio for total mortality. However, there could have been confounding factors for which they could not allow.

Most long-term studies show a disappointing waning of the initial HbA_1c improvement. Perhaps there is a case for educational updates. In a small trial with only 23 patients, Carlone et al. 67 randomised patients on long-standing pump therapy to standard care or to six educational weekly group meetings on advanced features of the pump, carbohydrate counting and other aspects of diet. After 6 months, the intervention group had reduced their HbA_1c by 1%. The control group did not change.

New developments

The main developments in pump therapy have been the use in combination with CGM systems, of which there are two forms that are now relevant to the pump. The first is when the CGM device is integrated with the pump, which means that it sends glucose results to the pump every 5 minutes or so, from a sensor just under the skin. Strictly speaking this means that the glucose result is for the level in interstitial tissue, not in the bloodstream, but the two are closely related. With the integrated CGM system, the pump can send alarms to the user, following which they can take action. This helps users to avoid hypoglycaemic episodes, but some find the alarms to be a nuisance and may disable the alarms. False alarms are not uncommon.

Four trials of CGM compared the pump with CGM against MDI with SMBG, which confounds things [Hermanides et al. (Eurythmic trial),68 Lee et al. ,69 Peyrot and Rubin,70 Bergenstal et al. (STAR-3)71]. The durations of these trials were only 6, 3.5, 3.7 and 12 months, respectively.

Continuous glucose monitoring would have implications for the use of pump therapy rather than MDI if the effectiveness of CGM differed between the two forms of treatment. Garg et al. 72 found that the effects of real-time CGM in reducing HbA_1c and hypoglycaemia were similar in two matched groups on MDI and pump.

More recently, the Medtronic Veo (Medtronic, Watford, UK) has been introduced, which has a facility to link with a CGM system and suspend insulin infusion (the LGS facility) if the glucose level goes too low, for up to 2 hours. This means that the pump can take action. In practice, most suspensions are for much less than 2 hours because the wearer takes action. However, at night when the wearer is asleep, this may not happen. 73

There have been two trials of the Veo suspend pump. In the ASPIRE (Automation to Simulate Pancreatic Insulin Response) trial74 in the USA and Canada, the recruits were familiar with pump therapy. They had a 2-week run-in period and were selected for the trial if they had nocturnal hypoglycaemia (defined as plasma glucose < 3.7 mmol/l) at least twice in that period. They also had to be prepared to wear the sensors at least 80% of the time. They were randomised to the Veo with its LGS facility, or to the Medtronic Paradigm Revel, which has integrated CGM but no LGS action. The trial was sponsored by Medtronic, and Medtronic staff were involved in data analysis and editorial assistance. 74 The trial showed no difference in HbA_1c after 3 months, perhaps not surprisingly because the baseline HbA_1c was very good at 7.2% or 55 mmol/mol. There was reduced hypoglycaemia, especially nocturnal. In the Veo group, 111 of 121 patients had at least one nocturnal suspension on the pump that lasted 2 hours. A 2-hour suspension does not lead to significant ketosis. QoL measures, EuroQol-5 Dimensions (EQ-5D) and the Hypoglycaemia Fear Survey (HFS) score, showed no difference between the arms of the trial. There were only four severe hypoglycaemic episodes, none in the Veo arm.

So the main benefit of the Veo LGS over the integrated CGM pump system is reduction of nocturnal hypoglycaemia. There is quite a large extra capital cost for the device (£2692) with an annual cost, including consumables, of £4862. This will make it difficult to prove cost-effectiveness. The group in which the Veo is most likely to be cost-effective will be patients with recurrent severe hypoglycaemia, but that group is covered by existing NICE guidance on the pump and is not recruited to the REPOSE Trial.

The other trial of the Veo was by Ly et al. 75 in Australia. This trial recruited mainly children and adolescents, with only 31% aged > 18 years. Patients were selected on the basis of impaired awareness of hypoglycaemia. They had been on pump therapy for an average of 4 years. They were randomised to the Veo suspend pump or to stay on their previous pump and use SMBG – not CGM. This immediately raises a problem because the Veo arm has both the LGS facility and CGM. It would have been better to have CGM in both arms. A more serious problem with the study is that, despite reasonable numbers (49 to pump plus SMBG, 45 to the Veo) and randomisation, there was a very marked baseline mismatch in previous severe (defined as seizure or coma) and moderate (defined as requiring assistance) hypoglycaemia, with a rate of 130 per 100 patient-months [95% confidence interval (CI) 111 to 150 patient-months] in the Veo group, but only 21 per 100 patient-months in the control arm (95% CI 14 to 30 patient-months). At study end after 6 months, the rate of moderate and severe hypoglycaemic episodes was 28.4 in the Veo group and 11.9 in the control arm. However, these figures were reversed when the authors adjusted for baseline rates, from 28.4 to 9.5, and from 11.9 to 34.2, all per 100 patient-months. There were no significant changes in HbA_1c, but both groups started with quite reasonable levels of 7.6% and 7.4%.

The analysis by Ly et al. 75 has been strongly criticised by the German Institute for Quality and Efficiency in Healthcare [Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG)], as reported by Heinemann and Hermanns. 76

The Veo has been appraised by the NICE Diagnostics Assessment Programme. Their guidance is shown in Box 1.

The patient group in which it has been recommended is different from that in the REPOSE Trial, and so the arrival of the Veo and its LGS facility has no implications for the implementations of the results of the REPOSE Trial.

Findings: structured education

The DAFNE course has changed little since the original trial published in 2002. 2 A programme of work has included a trial comparing the 5-day course in 1 week with 1 day a week for 5 weeks, which found little difference in outcomes. 78

One finding from the DAFNE research programme has been that many patients doing the DAFNE course, in preparation for going on to pump therapy, no longer need to progress to a pump after completing the course. Ray et al. 24 reported that after DAFNE education, 69% of patients previously being considered for pump therapy could remain on MDI.

However, another study from the programme (Mansell et al. 79) found that some patients who had been through DAFNE education still benefited from pump therapy in terms of a reduction in stress [measured by the PAID (Problem Areas in Diabetes Questionnaire) score] and improved glycaemic control at 12 months of follow-up. This may have been because individuals who progress to the pump after doing a DAFNE course have higher pre-course stress levels. 80

Conversely, attendance at DAFNE courses sometimes identifies individuals for whom pump therapy is indicated because of a troublesome dawn phenomenon.

Research into DAFNE education has also been undertaken by the Irish DAFNE group. 81 They carried out a large randomised trial of group follow-up compared with individual clinic visits for patients who had completed the DAFNE course. The intervention group received group education at 6 and 12 months after DAFNE, following a semistructured curriculum, whereas the control group had individual clinic appointments with doctor, nurse or dietitian. The additional education conferred no benefit over individual clinic visits.

The Irish DAFNE group also carried out a qualitative study to identify factors that influenced how well DAFNE graduates incorporated what they had learned into long-term daily living. 82 They identified four themes:

1. Being empowered, and feeling able to manage their diabetes, which some people did not manage to do.

2. Embedded knowledge, which increased over time, for example as patients got better at carbohydrate counting.

3. Maintaining motivation, including coping with uncertainty. The researchers commented that this was most marked at 6 months but improved later. Reducing the risk of complications was a strong motivation factor.

4. Continued support from health-care professionals.

The Australian OzDAFNE group83 also looked at psychological changes after the DAFNE course, and found increases in what they called ‘mastery/control’ and a reduction in diabetes-related distress. One of their key points was that the mean duration of diabetes in their participants was 18 years, but they had low self-assessment of their ability to manage their diabetes, so they recommended that referral to DAFNE courses should not be restricted to recently diagnosed patients.

Findings: new insulins

Some new basal insulins have been introduced, including degludec (Tresiba, Novo Nordisk, Gatwick, UK) and glargine 300 (Toujeo, Sanofi-Aventis, Guildford, UK). However, these are very long-acting basal insulins, and may not have the flexibility in dosing that is needed in MDI for T1DM, and with no data to date about how these might be used effectively in patients with T1DM who are undergoing structured education.

Newer short-acting insulins include ‘fast aspart’, which, in pump therapy, is reported to have a faster glucose-lowering effect but with the same effect overall. The implications for glycaemic control and hypoglycaemia were not reported by Zijlstra et al. 84

Findings: quality of life

Past reviews found a disappointingly low amount of evidence on QoL. This has implications for cost-effectiveness analysis. The Thomas et al. 15 pilot trial of pump therapy versus analogue MDI reported QoL as measured by the Diabetes Quality of Life questionnaire (DQOL) but found no difference. With only seven patients in each arm this may not be surprising. The 2008 health technology assessment (HTA)9 for NICE identified 48 observational studies of pump therapy, but only one reported QoL in adults, and it was a before-and-after study in which patients switched to the pump from conventional insulin therapy, not analogue MDI.

One observational study85 published since then has compared QoL. This study85 by the EQuality1 Study Group from Italy, has both strengths and weaknesses. It was a very large case–control study, with 1341 people with T1DM from 62 clinics, with 481 on the pump and 860 on MDI. The MDI patients came from centres both with and without pump services. Reliable instruments were used: the diabetes-specific quality of life scale (DSQOL) for QoL, Diabetes Treatment Satisfaction Questionnaire (DTSQ) for treatment satisfaction and Short Form questionnaire-36 items (SF-36) for health status. Eighty-four per cent of patients on the pump had been on it for > 1 year; 90% of the MDI group were on glargine-based MDI with the rest using NPH. All of the MDI patients had been on at least four insulin injections a day for > 6 months. The pump and MDI groups were well matched on some variables, but there were striking differences in carbohydrate counting (56% of pump group vs. 40% on MDI) and self-adjustment of insulin doses (80.5% vs. 66.5%), suggesting a marked educational imbalance.

Some DSQOL results were slightly better among the pump group, but this was statistically significant only for diet restrictions (65.5 vs. 60.8; p = 0.0003). DTSQ scores were better on pump (30.2 vs. 26.2; p < 0.0001). SF-36 scores were better on MDI, but in most domains, not statistically significantly so. However, the authors report that multiple regression analysis (details not provided, but adjusted for clinical factors including complications, which were more common in the pump group) showed that the pump group had significantly better scores in DSQOL diet, daily hassles and fear of hypoglycaemia. No differences were found between NPH and glargine-based MDI. The study was supported by Medtronic.

The lack of difference between the QoL effects of NPH and glargine-based MDI may not apply to detemir (Levemir, Novo Nordisk, Gatwick, UK)-based MDI, because detemir given twice daily may provide a more flexible lifestyle than once-daily glargine.

The Five Nations Study43 was a good-quality trial completed before long-acting analogues became available. It compared pump therapy with lispro and NPH-based MDI, but, unusually, the NPH insulin could be given up to four times a day and only 41% of patients had it once daily, with 32% getting it twice a day and 23% thrice daily. The study reported QoL using DQOL and Short Form questionnaire-12 items (SF-12). With SF-12 there was no difference in physical state but the pump group did better on mental health. The end of trial DQOL was 75 on pump and 71 on MDI, a small but statistically significant difference (p < 0.001). The difference reflected gains in treatment satisfaction, flexibility of eating and lifestyle, and reduced worry.

Study design

The REPOSE Trial was a pragmatic, multicentre, parallel-group, open-label, confirmatory cluster RCT. Participants were allocated a place on a week-long DAFNE course, depending on their availability to attend the course. The course (cluster element) groups were then randomly allocated in pairs to either pump or MDI treatment, with allocation concealed. A cluster design was chosen because of the impracticality of randomising individuals and then finding suitable times for that participant to attend a course of the correct allocation. 23 Such an approach was more likely to have resulted in significantly higher attrition rates pre course. Following the course, participants received the trial treatment for 2 years and outcome measures were collected at 6, 12 and 24 months post course. Outcome measurement was not blinded (see Data collection).

Approvals obtained

The protocol was approved by the Research Ethics Committee (REC) North West, Liverpool East, on 26 April 2011 (REC reference number 11/H1002/10). Each participating centre gave UK NHS Research and Development (R&D) approval (see Appendix 2). The protocol received Medicines and Healthcare products Regulatory Agency (MHRA) clinical trials authorisation on 26 May 2011 [European Union Drug Regulating Authorities Clinical Trials (EudraCT) reference no: 2010-023198-21].

Setting

The trial was conducted in eight secondary care diabetes centres in Sheffield, Cambridge, Dumfries and Galloway, Edinburgh, Glasgow, Harrogate, London and Nottingham (see Table 11). Participating centres all had experience in delivering high-quality structured education using DAFNE and had variable levels of experience delivering pump therapy; most were established pump centres but some were relatively new to pump therapy. Nottingham was a reserve centre, activated midway through the trial. The seven centres involved from the outset were asked to recruit 40 participants to three pump and three MDI courses (5–8 patients on each course) over 11 months. Owing to a higher than anticipated dropout rate prior to the DAFNE courses we then recruited to an additional pair of courses at Harrogate, and a pair of courses at the reserve centre, Nottingham.

Participants

Participants were eligible for the trial if they met the following inclusion criteria:

1. were aged ≥ 18 years

2. had T1DM for at least 12 months at the time of the DAFNE course

3. were fluent in speaking, reading and understanding English

4. were willing to undertake intensive insulin therapy with SMBG, carbohydrate counting and insulin self-adjustment

5. had no preference for either pump or MDI, and were happy to be randomised

6. were currently using, or willing to switch to, insulin detemir

7. had a need for structured education to optimise diabetes control.

Furthermore, participants were excluded if they met any of the following criteria:

1. had already completed a diabetes education course

2. used a pump in the previous 3 years (defined as > 2 weeks’ use in the last 3 years) or had strong clinical indications for pump therapy in the view of the investigator

3. had renal impairment with a chance of needing renal replacement therapy within the next 2 years (enrolment staff to check that creatinine levels not > 200 µmol/l).

4. had uncontrolled hypertension (diastolic blood pressure of > 100 mmHg and/or sustained systolic level of > 160 mmHg)

5. had a history of heart disease within the past 3 months

6. had severe needle phobia (severity of phobia assessed, considering if the phobia might preclude full participation in either treatment arm or influence the participant’s preference for pump therapy)

7. had a current history of alcohol or drug abuse

8. had serious or unstable medical or psychological conditions that are active enough to preclude the participant safely taking part in the trial (based on investigatory judgement)

9. had recurrent episodes of skin infections

10. were pregnant or planning to become pregnant within the next 2 years

11. had taken part in any other investigational clinical trial during the 4 months prior to screening

12. had any other issue that might have precluded them from satisfactory participation in the study based on investigatory judgement

13. were unable to give informed consent.

Interventions

Primary outcomes

The main primary end point was the change in HbA_1c at 24 months, in those participants whose baseline HbA_1c was ≥ 7.5% (58 mmol/mol). The key secondary end point was the proportion of participants reaching the NICE target of a HbA_1c level of ≤ 7.5% (58 mmol/mol) at 24 months (of all participants).

Glycated haemoglobin is the accepted gold standard measure of glycaemic control and provides a measure of efficacy. Most health economic models of T1DM estimate the cost-effectiveness by primarily modifying HbA_1c levels, which subsequently affect the risk of diabetic complications. 88 However, it is important to note that HbA_1c may not have fallen in patients who entered the trial with low baseline levels of HbA_1c, but who might have been experiencing frequent hypoglycaemia or wished to increase dietary freedom. Success for such individuals would be a HbA_1c level that is maintained, or even rises slightly, with a reduction in the frequency of hypoglycaemia. 23 We included such patients as they could provide important information about QoL and the potential of pump therapy to reduce rates of hypoglycaemia. However, as their glycaemic control may not alter, including their HbA_1c data would have reduced our statistical power to establish improvement in our primary end point. We therefore powered the trial on the number of participants with a baseline HbA_1c of ≥ 7.5% (58 mmol/mol) and in whom a fall would reflect a worthwhile improvement in glycaemic control. We ensured standardisation by testing HbA_1c in a central laboratory.

Secondary outcomes

Secondary outcomes were evaluated in all participants and were measured at 6, 12 and 24 months. Blood and urine samples for secondary outcomes were tested in local laboratories.

Sample size

It is generally accepted that a difference of 0.5% (5.5 mmol/mol) in HbA_1c is clinically worthwhile. To detect this difference with a standard deviation (SD) of 1% at 80% power and 5% two-sided significance using a t-test requires 64 patients per group, for subjects > 7.5% HbA_1c. To allow for a clustering effect of the educators, with an average of seven patients per DAFNE group and a within-course intraclass correlation coefficient (ICC) of 0.05, common in diabetes care, the sample size increases to 84. Allowing for a 10% dropout over 24 months, the sample size per group becomes 93. Audit of the DAFNE database showed us that 75% of subjects had a HbA_1c of ≥ 7.5%, therefore requiring 124 subjects per group and 248 in total. We planned to recruit 280 subjects, which increased the power to 85% but allowed for some variation in dropout rates and the proportion of patients with HbA_1c ≥ 7.5%. However, monitoring of baseline data showed that the actual proportion of participants with HbA_1c ≥ 7.5% was around 90% rather than 75%. A modelling exercise undertaken during recruitment, with conservative estimates of 85% (HbA_1c ≥ 7.5%) and dropout rate of 15%, suggested that the trial would require at least 240 participants with primary outcome data at 2 years in order to preserve power of at least 85%. 23

Recruitment

A number of methods were used to approach potential participants:

• PIs or educators identified people from DAFNE waiting lists. They then telephoned or wrote to potentially eligible individuals.

• Individuals attending a clinic appointment with a trial PI or educator were offered the option of a future or immediate consultation regarding the trial.

• Clinicians [general practitioner (GP), dietitian, nurse] provided information to patients and referred them to PIs to be screened and enrolled.

• Details of the trial were advertised through the use of posters and leaflets in clinics (diabetes outpatient, dietetic, GP surgery).

• Reception staff in diabetes clinics were informed about the trial and provided with leaflets to give to patients who expressed an interest.

• Participant identification centres were used at some research centres to assist in the identification of suitable participants.

Interested individuals were given the opportunity to discuss the trial with the PI or educator. Those who were still interested in taking part were screened for eligibility. Those who were eligible were either invited to attend a local information meeting, at which the trial was discussed in detail and questions answered, or were provided with a patient information sheet and consent form and given the opportunity to ask further questions. Individuals who were still wanting to take part consented to the trial by one of three methods: (1) by returning a completed consent form (see Appendix 3) in the post, (2) by completing the form with the PI or educator or (3) by completing the form at a local information meeting. The participants’ contact details, GP details and ethnicity were also collected.

Allocation to Dose Adjustment For Normal Eating courses and randomisation

Following consent, participants were allocated to a REPOSE DAFNE course, depending on the participants’ availability. 23 Up to eight participants were allocated to each course, with a minimum of five preferable. Courses were randomised, in pairs, to either DAFNE with pump or DAFNE with MDI. 23 Participant allocation to courses was finalised for each course pair before randomisation took place, no less than 6 weeks prior to the date of the first DAFNE course in that pair. For the first seven centres, a simple randomisation procedure in block size of ‘2’, stratified by centre, was used for courses 1–4. Courses 5 onwards were allocated in pairs using minimisation of the overall and number of participants, with most recent baseline HbA_1c value of ≥ 7.5% or < 7.5% between the treatment groups. Any additional courses were allocated using minimisation. Known dropouts prior to the DAFNE course were excluded from the minimisation algorithm for future course allocation. A validated user-written Stata^® 13 (StataCorp LP, College Station, TX, USA) code was produced to generate the allocation by a statistician within Sheffield Clinical Trials Research Unit (CTRU), who implemented the randomisation. The trial co-ordinator revealed the allocation to study centres. 23

Blinding of the course allocation was not possible because of the nature of the treatment. Course allocations were revealed to centres 4–6 weeks prior to the date of the first course to allow sufficient preparation time. Participants were informed of the allocation of their DAFNE course no earlier than 4 weeks prior to that course. At this point they were asked to keep a record of any new episodes of moderate hypoglycaemia, which would be collected at the baseline assessment. If the course was a pump course, the participant was booked into a pre-course pump session, up to 3 weeks prior to the course date, in addition to the baseline assessment, which had to take place before the pump session.

If, for any reason, participants were unable to take part in the course at short notice, they could be allocated to a later course date, but only in the same trial arm as in the course to which they were originally allocated. Centres could also keep a list of reserve participants for courses, agreed prior to the time when the course allocation had been revealed to the educators. In the case of participants dropping out, the next person on the reserve list would be invited to participate in that course.

Data collection

Study visits took place at the participants’ diabetes centre. A data collection form (DCF) (see Appendix 4) was completed by the educator with the participant. Blood and urine samples were taken and analysed at local laboratories. Two blood samples were taken for measurement of the primary outcome (HbA_1c). One of these was analysed at a central laboratory as the primary measure and the second was tested at the local laboratory as a back-up. DCF data were entered at local centres on to the in-house Prospect web-based electronic data capture system, managed by the CTRU.

Baseline assessments took place up to 3 weeks prior to the DAFNE course. The educator completed the DCF with the participant and handed him/her the self-complete psychosocial questionnaire, asking for return of the completed questionnaire at the forthcoming DAFNE course. Additional demographic data collected at baseline were date of birth, sex, qualifications (highest qualification obtained) and current occupation. Participants were also handed a SAE contact card to aid in contacting their diabetes centre in the event of an AE.

At the DAFNE course, an attendance form was completed, detailing any missed sessions. The completed baseline psychosocial questionnaire was collected and the baseline DCF moderate hypoglycaemic episodes section was updated so that a full 4 weeks of hypoglycaemic episodes were recorded. At all time points, psychosocial questionnaires were posted from centres to Sheffield CTRU and entered on to Prospect by Sheffield CTRU clerical staff.

Participants were followed up at 6, 12 and 24 months after the DAFNE course. Participants were sent the blood glucose diary (see Appendix 5) and instructions for recording moderate hypoglycaemic episodes 4 weeks prior to each visit. Additionally, participants were posted the self-complete psychosocial questionnaire pack prior to the visit and asked to bring their completed questionnaire to the appointment, along with the blood glucose diary and record of moderate hypoglycaemic episodes.

Severe hypoglycaemic episodes or SAEs were collected from participants if reported over the telephone or in clinic. Any additional diabetes-related contacts (DRCs) were also recorded (see Appendix 6 for ongoing data collection booklet).

Blinding of outcome measures was considered impractical because of the intervention-specific nature of outcome measures and the necessity of a local diabetes nurse to collect the data. However, use of an objective outcome (HbA_1c) measured in a central laboratory will have minimised bias on the primary end point.

Trial completion

Participants were deemed to have completed the study if they had trial data recorded at baseline and 24 months. Participants were withdrawn from the study if:

• The participant asked to fully withdraw from the trial. On requesting withdrawal from the trial, participants were able to consent to continue to have their routine HbA_1c results recorded.

• The participant died.

Participants who were changing treatment continued in the trial unless formally withdrawn. Participants were deemed lost to follow-up if they failed to attend the baseline visit, DAFNE course or 24-month follow-up.

Research governance

The trial sponsor was Sheffield Teaching Hospitals NHS Foundation Trust. The trial was conducted in accordance with Good Clinical Practice (GCP) and the Medicines for Human Use (Clinical Trials) Regulations 2004. 92 All staff recruiting participants to the trial had undertaken GCP training. In line with the three-level categorisation of clinical trial risk in the Medical Research Council/Department of Health (DH)/MHRA report on risk-adapted approaches to the management of clinical trials of investigational medicinal products93 (based on the classification by Brosteanu et al. 94), the REPOSE Trial was classified as a Type A study: no higher than the risk of standard medical care. The trial treatment in REPOSE was licensed and administered according to its market authorisation. Trial-specific labelling was not used. Given the lack of criticality of the investigational medicinal product (IMP) with the data analysis and trial results, and the design of the trial being equivalent to standard care, there was no IMP tracking and accountability undertaken.

Three committees were established to govern the conduct of the study: an independent Trial Steering Committee (TSC), an independent Data Monitoring and Ethics Committee (DMEC) and a TMG. Full membership of the TSC and DMEC are listed at the end of this report. The committees functioned in accordance with Sheffield CTRU standard operating procedures (SOPs). The TSC was responsible for overall supervision and monitoring of the trial; it considered any recommendations from the DMEC and provided advice on any actions to be taken. The DMEC operated within a charter agreed by all members and was responsible for monitoring efficacy and safety data. Any concerns were reported to the TSC with recommendations. The TMG was responsible for supporting the implementation of the trial.

Reporting of adverse events

Adverse events were defined as any untoward medical occurrence in a participant to whom a medicinal product has been administered, including occurrences that are not necessarily caused by or related to that product. SAEs were defined as any AE that results in death; is life-threatening (subject at immediate risk of death); requires inpatient hospitalisation or prolonging existing hospitalisation; results in persistent or significant disability or incapacity, or consists of congenital anomaly or birth defect; or is another important medical event that may jeopardise the participant. Pregnancy was also recorded as a SAE, so that any AEs could be identified if and when the child was born.

Included as AEs were an increase in frequency of hypoglycaemia, a blood glucose reading > 30 mmol/l, unexplained constantly raised blood glucose readings, suspicion of pump malfunction and pump site infection. Excluded as AEs were non-serious episodes of hypoglycaemia and ketonuria.

Details of AEs were collected during follow-up appointments. Participants were also provided with a contact card and encouraged to get in touch with their diabetes team if they had experienced any adverse health events. SAEs were reported in accordance with the Sheffield CTRU and REPOSE SAE SOPs. SAEs were assessed by the local PI and reported to Sheffield CTRU within 24 hours of becoming aware of the event, with the exception of events that had been stated as exempt from immediate reporting, for which 28 days was allowed. These exemptions were episodes of severe hypoglycaemia requiring hospitalisation, episodes of DKA and pregnancy. SAEs were assessed for seriousness, frequency, intensity, relationship to study product and, when applicable, relationship to pump. The Summary of Product Characteristics for NovoRapid and Levemir (Novo Nordisk, Gatwick, UK) were kept on file as the reference safety information for the assessment of events. AEs were reviewed at regular intervals by the three study oversight committees. The chief investigator and DMEC chairperson were notified of all SAEs on the event being reported.

Reporting of protocol non-compliances

Protocol non-compliances were reported and assessed in accordance with the Sheffield CTRU and REPOSE non-compliances SOPs. A non-compliance was defined as ‘a departure from the protocol or GCP that has been identified retrospectively’. Non-compliances were addressed with staff training or, when appropriate, an amendment to the protocol. In line with MHRA guidance, deliberate prospective protocol non-compliances or ‘waivers’ were deemed to be unacceptable. A prospective list of exemptions from reporting and of pre-specified major and minor non-compliances was drawn up by the CTRU, the chief investigator and the sponsor.

Trial monitoring

Responsibility for monitoring was delegated to the CTRU and conducted in accordance with CTRU SOPs. Both on-site and central monitoring methods were adopted. Onsite monitoring visits took place at all centres at study set-up, prior to delivery of the first DAFNE course, post delivery of DAFNE course 2 and at study closeout. A further monitoring visit took place during follow-up at seven centres. At each visit, the study site file and key essential logs were reviewed for completeness. Source data verification was conducted for 100% of consent and SAE forms. Patient hospital records were reviewed to substantiate participant existence and eligibility (for which criteria were verifiable from hospital records). Monitoring reports were issued after each visit detailing any remedial actions required. Central monitoring tasks included point of entry validation, verification of data and post-entry validation checks. One participant per DAFNE course per centre was randomly selected for verification. Case report forms at all data collection time points were reviewed for completeness and quality, and verified to monitor data entry. Source data verification also took place for 100% of central laboratory HbA_1c results. Feedback on verification was provided and additional verification was undertaken when concerns were identified.

Statistical methods

All statistical analyses were performed in Stata 13 onwards. The MDI is the reference group for all treatment comparisons.

Patient and public involvement

As part of our recent work funded by a NIHR programme grant (PGfAR),27 15 DAFNE graduates were recruited to act as a ‘user group’ and contribute to different aspects of the work. We invited two members to join both the steering group and other investigator meetings. In addition, one of the project team is a pump user. They provided input to the trial design, implementation and dissemination, including all participant materials. 23 The work supported by the programme grant included qualitative studies in which the barriers to self-management in T1DM were explored. This work led to the development of a pilot study within the PGfAR work, in which a modified DAFNE course incorporating a pump curriculum was developed and piloted in three centres.

Aim

To ensure that there was consistency in:

1. the delivery of the 5-day DAFNE pump curriculum

2. the timing and content of pump pre-assessment/setting up on pump session.

Multiple daily injection courses were not included in the FT, as there exists a rigorous quality assurance programme of MDI courses in standard care, and trial centres are routinely audited.

Methods

An experienced DAFNE educator and peer reviewer from Sheffield Teaching Hospitals NHS Foundation Trust was employed as the fidelity assessor (FA); this educator was not directly involved in the delivery of REPOSE courses. The FA assessed whether or not the pump courses delivered the correct DAFNE content and philosophy. The FA visited each centre to observe the ‘Wednesday’ of the pump DAFNE course. Wednesday was chosen as pump curriculum sessions that incorporated key differences to MDI would be delivered by DAFNE educators from both dietetic and nursing specialties. In addition, patients on the course should have settled into the course, be more relaxed and be starting to establish patterns and adjustments to their regimen by the third day. It was planned that the FT take place on the first or second pump course at each centre.

Experienced educators who devised the pump curriculum and the national director of the DAFNE programme discussed which sessions differed most between the pump and MDI DAFNE curricula and, thus, warranted observation. These sessions were decided as follows:

• daily goals, blood glucose results and insulin doses

• insulin dose adjustment theory, basal rate testing

• dose adjustment practice – reducing and increasing insulin

• setting up Bolus Wizard

• sick day rules

• alcohol

• exercise.

All but one session was scheduled for observation on the FT visit, as it was not possible to timetable all sessions that differed between the MDI and pump courses on 1 day. In lieu of observation, the FA reviewed the lesson plan for the sick day rules session.

The following data and documents were requested to be made available for the FT visit:

• pre-course pump session details including patient attendance, session timings and lesson plan

• pump course timetable

• list of course participants and details

• lesson plan for all observed sessions and the sick day rules lesson plan.

A DEP peer support learning outcomes form was completed for each session observed. This form listed the essential learning outcomes for each session and the FA evaluated whether or not these were met (or partially met). For each learning outcome, the FA provided evidence for its achievement. A template report was devised and used to collate the data collected from the FT visit.

Once the FA had completed the assessment, feedback was given immediately so that educators could resolve any problems. The report was completed within 3 days of the visit and sent to the trial management office.

Setting and perspective

The health economic analyses are designed to inform UK decision-makers within the UK NHS on the potential resource implications of choosing to use pump therapy with DAFNE structured education (pump + DAFNE) or MDI with DAFNE structured education (MDI + DAFNE) for the group of adults with T1DM in the REPOSE Trial, comprising adults with T1DM who are naive to pump therapy.

To ensure that all economic analyses were applicable to the UK decision-making setting, all economic analyses took the NHS and Personal Social Services perspective in line with (NICE) guidance. 22

Two approaches: economic evaluation alongside the clinical trial and long-term cost-effectiveness modelling

The cost-effectiveness of ‘pump + DAFNE’ compared with ‘MDI + DAFNE’ was assessed using an Economic Evaluation Alongside Clinical Trials (EEACT) and long-term modelling exercise. The EEACT took a 2-year time horizon and the long-term modelling took a lifetime horizon. As the long-term modelling takes a lifetime time horizon, and includes all clinically important complications of diabetes, this should be considered as the primary analysis.

Price year and discounting

All costs are reported in 2013–14 prices; if costs were obtained from a previous financial year they were inflated to 2013–14 prices using the Hospital and Community Health Services Pay and Prices index. 96 All costs and QALYs were discounted at a rate of 3.5% in line with NICE guidance. 22 All costs and QALYs were assumed to fall at the end of the year, apart from the cost of the structured education courses, which were assumed to occur at the start of the first year.

Population and subgroups for analysis

The individuals in the REPOSE Trial were adults with T1DM who were eligible to receive a structured education course. Furthermore, all individuals were naive to insulin pump therapy and did not have a preference to receive the pump. The average age of participants was 40.4 years and their mean duration of their diabetes was 18.0 years. Data were collected from individuals at baseline and at 6 months, 1 and 2 years post randomisation. In the MDI + DAFNE arm, 6, 3 and 5 individuals out of 135 were lost to follow-up at 6 months, 1 and 2 years, respectively. A further individual in the MDI + DAFNE arm withdrew from the trial at 6 months. In the pump + DAFNE arm, 0, 1 and 2 individuals out of 132 were lost to follow-up at 6 months, 1 and 2 years, respectively. A further individual in the pump + DAFNE arm withdrew from the trial at 1 year.

The data collected in the REPOSE Trial were considered to be the only relevant evidence on the relative effectiveness of pump + DAFNE compared with MDI + DAFNE. This is because REPOSE is the only large study in a UK setting in which adults with T1DM in both trial arms have received equivalent diabetes education in both the pump and MDI trial arms.

There are two populations in the REPOSE Trial: (1) the ITT population and (2) the per-protocol population. The ITT population includes all individuals who graduated their DAFNE course and had follow-up data for at least one data collection period. In the ITT population, individuals were assigned to their randomised treatment irrespective of whether or not they switched to the other insulin delivery mechanism. The per-protocol population includes all of the individuals who were in the ITT population and adhered to their insulin delivery mechanism (either pump or MDI). Unless otherwise stated, all analyses of the REPOSE Trial data to inform the health economic analyses were conducted in the ITT population.

The population analysed in the primary health economic analyses is all individuals in the REPOSE Trial, regardless of whether or not the individual’s baseline HbA_1c was ≥ 58 mmol/mol (7.5%). The analysis population differs from the population used in the primary clinical end point statistical analysis, as the base-case health economic analysis focuses on the whole trial population rather than those individuals with a HbA_1c of < 58 mmol/mol (7.5%). For the health economic analyses, it is important to assess the cost-effectiveness of pump + DAFNE compared with MDI + DAFNE for all adults with T1DM who would be potentially eligible to receive either treatment if they were adopted as standard practice.

Subgroup analyses 1–6 were conducted in the long-term modelling only, because of concerns about the reduction in sample size potentially producing spurious results in the EEACT. However, subgroup analysis 7 was conducted in the EEACT, as this was an important subgroup analysis for the estimation of treatment effect of HbA_1c (see Statistical methods). The subgroup analyses were conducted in following subgroups of the REPOSE Trial participants:

1. baseline HbA_1c ≥ 58 mmol/mol (7.5%)

2. baseline HbA_1c ≥ 58 mmol/mol (7.5%) and < 69 mmol/mol (8.5%)

3. baseline HbA_1c ≥ 69 mmol/mol (8.5%) and < 80 mmol/mol (9.5%)

4. baseline HbA_1c ≥ 80 mmol/mol (9.5%)

5. baseline HbA_1c < 69 mmol/mol (8.5%)

6. baseline HbA_1c ≥ 69 mmol/mol (8.5%)

7. all individuals in the per-protocol population.

Cost of the Dose Adjustment For Normal Eating course

A detailed within-trial costing of the DAFNE courses was not undertaken because DAFNE is an already established intervention within the NHS. The cost of DAFNE training for adults with T1DM using MDI has been calculated by DAFNE UK as being £359.10 per course attendee in 2012–13 prices (£363.10 in 2013–14 prices). 97

Based on discussion with experts involved in the REPOSE Trial, including a Professor of Clinical Diabetes and Honorary Consultant Physician, and a Professor in Public Health and Health Technology Assessment, it was assumed that the cost of a DAFNE course in the pump + DAFNE arm is identical to the cost of a DAFNE course in the MDI + DAFNE arm, except for the cost of staff time spent conducting an additional pre-course pump-fitting session.

Data were collected on the time spent delivering a pre-course fitting session for pump + DAFNE participants in the FT process. The FT process was conducted for one pump + DAFNE course at each trial centre to ensure that the pump + DAFNE course taught the principles of insulin adjustment in a similar fashion to the MDI + DAFNE course. These data were utilised to estimate the additional cost of the pre-course pump fitting session in the pump + DAFNE arm. Expert advice was sought from two centres in which it was unclear whether reported time use as part of the FT referred to the educator time spent or the total time individuals spent at the venue (which could include non-contact waiting time). To ensure consistency between the estimated costs of a MDI + DAFNE course and a pump + DAFNE course, the cost of staff time was obtained from the estimated cost of staff time for the MDI + DAFNE course. The cost of the pre-course fitting session was estimated to be £28.82 per adult with T1DM.

Economic analysis alongside the clinical trial of pump + Dose Adjustment For Normal Eating versus multiple daily injection + Dose Adjustment For Normal Eating

Long-term cost-effectiveness

The Sheffield Type 1 Diabetes Policy Model Overview

The Sheffield Type 1 Diabetes Policy Model (henceforth, the Model) was used to estimate the lifetime costs and QALYs for individuals receiving MDI + DAFNE and pump + DAFNE. The Model has been developed and used over several years, and a detailed description is provided in a journal article108 and a detailed report to the NIHR on the DAFNE programme grant research. 27 In this analysis, we have updated some aspects of the evidence used within the Model. We term the version used here as ‘The Sheffield Type 1 Diabetes Policy Model version 1.3’.

The Model is an individual-level simulation model, which consists of a series of submodels that simulate the progression of diabetic complications (microvascular and macrovascular), SAEs (severe hypoglycaemia and DKA) and mortality in a given population with T1DM. Each of the modelled microvascular (nephropathy, neuropathy, retinopathy and macular oedema) and macrovascular complications (MI, stroke, HF and angina) are included in the model as separate Markov submodels with an annual time cycle. Short-term AEs (severe hypoglycaemia and DKA) are modelled as the annual incidence of these complications, dependent on each patient’s characteristics. The Model structure is also presented in Figure 1. The Model attaches utilities and ongoing costs to health states and one-off costs to events (the move to another health state in a submodel). These costs and utilities are combined with the length of time that a patient spends in a health state to estimate lifetime costs and QALYs. The Model estimates patient’s disease progression over their lifetime.

FIGURE 1.

The structure of the Sheffield Type 1 Diabetes Policy Model (from Thokala et al. 108 with permission from John Wiley & Sons, Inc.). CVD, cardiovascular disease; ESRD, end-stage renal disease; hypos, hypoglycaemic episodes; PAD, peripheral arterial disease.

The disease progression parameters in the Model were not updated in these analyses. However, the costs and utilities associated with health states and events were updated. A full probabilistic sensitivity analysis (PSA) was conducted with 500 probabilistic runs, each with 5000 individuals in each Model arm. All Model runs were conducted using the SIMUL8 2010 professional (Simul8 Corporation, Boston, MA, USA) programme.

Aims and objectives

As noted in Chapter 2, evidence on QoL effects of the pump has been inconsistent, with some studies reporting no difference between the pump and MDI groups, and others reporting improved QoL on the pump. A previous HTA report identified some gains in QoL that could be described as ‘social related’ rather than ‘health related’. 14 These included flexibility of lifestyle and fewer problems dealing with variations in daily life, such as timing of meals. For this reason, we included a range of psychosocial measures alongside embedded qualitative research in the REPOSE Trial.

The psychosocial study employed a mixed-methods quantitative (questionnaires) and qualitative (interviews) approach to:

1. establish whether or not, and why, there were any differences in QoL and other psychological or psychosocial outcomes between participants using pump and MDI regimens

2. examine whether or not, and why, QoL and other outcomes changed over time

3. understand and explore the added benefit (if any) of pump technology over MDI from participants’ and educators’ perspectives

4. explore why some patients may do better than others using the pump

5. examine acceptability of, and reasons for, discontinuing (pump) treatment

6. enhance understanding and assist in the interpretation of trial outcomes.

Quantitative methods

Validated and reliable questionnaires were used to assess generic and health-specific QoL, treatment satisfaction, fear of hypoglycaemia, hypoglycaemia unawareness, self-efficacy, social support, adherence to treatment, emotional well-being and acceptability of technology. A repeated-measures longitudinal questionnaire study explored both differences in outcomes between the two trial arms and the short- and long-term predictors and mediators of outcomes. Outcomes were measured at baseline and at 6, 12 and 24 months after the DAFNE course. These time points were selected to capture both short- and long-term post-treatment changes in psychosocial outcomes. Questionnaires were posted to participants and self-completed within 6 weeks of the specified time point.

Qualitative methods

Sensitivity analysis: effect of centre and lead Dose Adjustment For Normal Eating course educator

We undertook a further analysis that used a nested model of patients within courses, which, in turn, are nested within course lead educators, to investigate differences in outcomes between lead educators. For this nested model, the ICC of the lower-level clustering variable, DAFNE course, is 0.5%; for the upper-level clusters, lead educator, ICC < 0.1%. We found no evidence of notable differences in outcomes between lead course educators. This analysis was performed for available data only.

We explored the centre effect through an interaction test between centre and treatment group. Results of estimated MD in HbA_1c change (% or mmol/mol) at 24 months are presented by centre with the aid of forest plots (Figure 8). The overall p-value for the interaction between treatment and centre was 0.565, suggesting that there is no centre effect. The centre with the largest difference between treatments was Nottingham, although the CI for this centre is large because of the small number of participants with outcome data at that centre.

FIGURE 8.

Forest plot of MD in 24-month HbA_1c change (%) at 24 months between groups for participants with baseline HbA_1c ≥ 7.5% by centre (complete cases with HbA_1c ≥ 7.5% at baseline, n = 224). MDs are calculated from mixed-effects regression analysis adjusted for DAFNE course. KCH, King’s College Hospital; MCID, minimum clinically important difference.

Glycated haemoglobin

The proportion of participants reaching the NICE target of HbA_1c of ≤ 7.5% (58 mmol/mol) after 2 years is displayed in Table 23 (including all participants regardless of baseline HbA_1c value). The proportion of patients with HbA_1c ≤ 7.5% was similar across the groups. The results are very similar at 6 and 12 months (Table 24).

Table 25 shows the distribution of HbA_1c categories at baseline and 24 months. Of the participants who ended with HbA_1c of ≤ 7.5% at 24 months, 12 began the study with baseline HbA_1c of ≥ 8.5%.

The primary analysis at 24 months displayed above (see Primary outcome) is repeated for 6- and 12-month follow-up visits among participants with complete data. The results for these interim follow-ups are consistent with the primary outcome analysis and are displayed in Table 26. The largest MD in HbA_1c change from baseline was observed at 6 months, –0.25% (95% CI –0.52% to 0.02%) or –2.7 mmol/mol (95% CI –5.6 to 0.2 mmol/mol), but is not clinically relevant or statistically significant at the 5% nominal level.

Episodes of moderate and severe hypoglycaemia

Few severe hypoglycaemic episodes were observed post baseline; 49 episodes recorded from 25 participants23 (Table 27). All severe hypoglycaemic episodes occurred while participants were on their allocated treatment. Across both treatment groups the number of severe hypoglycaemic episodes reduced: the average number of episodes per patient-year in the study reduced from 0.17 before baseline to 0.10 during follow-up. The IRR for the number of severe hypoglycaemic episodes in the 24-month follow-up, compared with the year before baseline, is 0.46 (95% CI 0.24 to 0.89; p = 0.021). 23 Therefore, compared with the year before baseline, the number of severe hypoglycaemic episodes per year were roughly halved in the 2 years of follow-up post baseline. There was no statistically significant difference in the rate of severe hypoglycaemia during follow-up between the treatment groups having adjusted for centre, DAFNE course, baseline HbA_1c and presence of at least one severe hypoglycaemic episode in the 12 months before baseline (IRR 1.13; 95% CI 0.51 to 2.51; p = 0.766). The comparison of severe hypoglycaemic episodes between groups was repeated excluding the first 6 months of follow-up, which is the ‘settling in’ period on the pump. This time the estimated IRR was almost equivocal, but the large CI around this reflects the amount of uncertainty as a result of these analyses being based on so few episodes from few participants (IRR 1.05; 95% CI 0.44 to 2.53; p = 0.912).

Across both treatment arms, on average, three moderate hypoglycaemic episodes were recorded per patient over a 4-week history at 6 months (Table 28). By 24 months, the average number of recorded moderate hypoglycaemic episodes during a 4-week history was slightly lower (2.6 for pump, 2.3 for MDI). There was no statistically significant difference between the groups in the rate of moderate hypoglycaemic episodes at any time point. 23

Few participants reported one or more severe hypoglycaemic episode during study follow-up: 14 (10.6%) in the pump group and 11 (8.6%) in the MDI group (Table 29). There was no evidence that the number of patients reporting at least on severe hypoglycaemic episode was different in the two groups: OR 1.22 (95% CI 0.49 to 3.03). More than half of the participants reported at least one moderate hypoglycaemic episode in the 4 weeks prior to follow-up at each time point and across both treatment groups. Slightly more participants reported at least one episode at 6 months in the pump group than in the MDI group (p = 0.088). However, a smaller proportion of participants reported episodes in the pump group at 12 and 14 months, although not statistically significant.

Figure 9 shows the distribution of the number of severe hypoglycaemic episodes for those who had one or more episodes post baseline. The majority of participants had only one severe hypoglycaemic episode, 10 participants recorded more than one episode during the follow-up period and the maximum recorded by a participant was seven. The number of patients who had an episode makes up a small proportion of the study population (10%).

FIGURE 9.

Distribution of number of severe hypoglycaemic episodes in participants with at least one episode post baseline, by treatment group. (a) Pump, n = 14; and (b) MDI, n = 11.

Figure 10 shows the timing of severe hypoglycaemic episodes. Each dot represents a severe hypoglycaemic episode. Dots connected by a line represent severe hypoglycaemic episodes experienced by the same person.

FIGURE 10.

Severe hypoglycaemic episodes over time per participant with at least one episode post baseline, by treatment group. (a) Pump, n = 14; and (b) MDI, n = 11.

There is no statistically significant difference in the odds of proteinuria between the treatment groups (Table 30). At 6 months, the odds of being in a higher proteinuria category (where macroalbuminuria is the highest category) are estimated to be 21% lower in the pump group than the MDI group (OR 0.79), but 14% higher at 12 months, and almost identical at 24 months.

Table 31 shows exploratory descriptive analyses of self-reported physical activity for the two groups at each study visit; no formal statistical tests have been performed on these data. The amount of physical activity appears similar across the groups.

Table 32 shows the results of comparing secondary continuous outcomes across treatment groups. Weight remained roughly constant throughout the study duration, and was not statistically significantly different between the treatment groups at any follow-up. 23 A slight increase in HDL cholesterol and a slight decrease in TC was observed across both treatment groups. 23 There was no evidence of a difference between treatment groups in cholesterol change from baseline, with p-values ranging from 0.219 to 0.856. Insulin dose decreased across both pump and MDI arms. There was evidence of a difference in the mean change in insulin dose at 12 months between treatment groups; on average, participants in the pump group had a 0.07-IU/weight larger decrease (95% CI 0.01 to 0.013 IU/weight; p = 0.017) in insulin dose than those in the MDI group. However, the difference between treatments in insulin dose was slightly smaller at 6 and 24 months, but not statistically significant.

Table 33 summarises blood glucose testing per day averaged over a 2-week recorded period, stratified by the baseline HbA_1c category. A post hoc analysis indicated that there was no difference in the mean blood glucose testing frequency between treatment groups at 24 months, having adjusted for baseline number of blood glucose tests, centre and DAFNE course. 23 The adjusted MD in blood glucose tests (95% CI) was 0.22 (–0.24 to 0.68) per day or 3.1 (–3.4 to 9.6) over 2 weeks; p = 0.352. Overall, the number of blood glucose tests increased from 3.6 at baseline to 4.1 per day at 24 months (95% CI 0.33 to 0.82; p < 0.001). 23

Adverse events

Table 38 shows the AEs recorded throughout study follow-up. More participants in the pump arm (66%) reported AEs than in the MDI arm (37%). However, part of this difference can be attributed to the 23 cases of suspicion of pump malfunction, which, by definition, could occur only for participants using pump therapy. Table 39 shows the AEs that were recorded over different time periods during study follow-up. During each time period, more participants in the pump arm experienced AEs than in the MDI arm. A total of 142 AEs were recorded for the pump group during the first 6 months of follow-up in comparison with 84 in the following 6 months, and 94 in the final 12 months of follow-up, suggesting that more AEs occurred during the early ‘settling in’ period on pump therapy.

Table 40 shows the AEs that were classified as being SAEs. The distribution of SAEs was similar across the treatment groups, with the exception that more participants experienced DKA in the pump group. Table 41 shows SAEs by study time period. Again, for the pump group, more SAEs were recorded in the first 6 months (n = 17) than in the following 6 months (n = 11) or when compared with the last 12 months (n = 17).

Note: All of the DKAs that occurred were reported as SAEs and resulted in hospitalisation. All of the SAEs have a corresponding AE; however, in some cases, a DKA SAE had a corresponding AE that was not labelled as DKA, which is why there are more DKA SAEs recorded than DKA AEs.

Characteristics of participants by missing data status

Tables 42 and 43 show baseline characteristics of patients with missing data.

Course characteristics

All eight REPOSE centres were fidelity tested. Four centres were fidelity tested on their second pump course and two centres were fidelity tested on their first pump course. One centre was fidelity tested on their third and final course as a result of personal circumstances of the FA precluding the assessment being undertaken on the second pump course. Nottingham ran one pair of courses and, thus, FT took place on its only pump course.

The number of REPOSE participants on the fidelity-tested pump courses ranged from 3 to 7 (for course sizes, see Table 44). The range of participants on the remaining pump courses was 3–8, with a mean of 5.7.

One pump course (Cambridge) included a non-REPOSE participant who had been on a pump for 10 years and was very keen to do DAFNE.

Pump pre-course session

All participants attended the pump pre-course session to learn the mechanics of pump therapy and to programme and load the pump with saline to enable practice and familiarisation prior to undertaking the course. This session was scheduled to run for 2 hours and 30 minutes (± 15 minutes). Seven out of the eight centres ran sessions within this duration window. The centre that did not (Nottingham) ran a pump pre-course session of 2 hours and so was 15 minutes short of the specified duration window.

The majority of centres delivered the pump pre-course session solely by REPOSE educators (diabetes specialist nurses and dietitians). Two centres (Glasgow and Edinburgh) had a Medtronic representative present to provide technical support and help with elements of pump set-up. One centre (Glasgow) also had the PI present.

All pump pre-course session lesson plans were evaluated by the FA as relating to the objectives set for this session.

Pump courses

The pump course timetable was reviewed by the FA. All centres provided timetables that were evaluated as incorporating all elements of the pump DAFNE curriculum in a logical order. Based on the times allocated for sessions on the pump course timetable, all centres planned to deliver the curriculum in the specified duration window of ≥ 1870 minutes but ≤ 2280 minutes. The mean course duration was 2006 minutes, that is 33 hours and 26 minutes.

All sessions planned to be observed were reviewed during the fidelity visit and their lesson plans were reviewed.

The sick day rules lesson plan was reviewed for each centre. Seven of the eight centres were evaluated as having no issues with this session lesson plan, with only minor problems noted, for example no aims or objectives listed, timings not written on. One centre (Glasgow) was evaluated as having an issue with the sick day rule lesson plan. The lesson plan was lifted directly from the pump DAFNE curriculum without personalisation. The Glasgow educator explained that there was no time to personalise the lesson plan but agreed to remedy for future courses.

Essential learning outcomes

The FA recorded (with evidence) if all essential learning outcomes were met in the sessions observed. Sessions were recorded as having met all learning outcomes: ‘yes’ or ‘no’ or partially achieving essential learning outcomes. Table 45 provides a summary.