Continuous subcutaneous insulin infusion versus multiple daily injections in children and young people at diagnosis of type 1 diabetes: the SCIPI RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 08/14/39. The contractual start date was in August 2010. The draft report began editorial review in May 2017 and was accepted for publication in September 2017. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Joanne Blair undertakes paid advisory work for, and has received funding for research, to attend academic meetings and to support a nursing salary from, Novo Nordisk Ltd (Gatwick, UK), a pharmaceutical company that manufactures some of the insulins used in the SubCutaneous Insulin: Pumps or Injections? (SCIPI) study. The work undertaken by Joanne Blair for Novo Nordisk Ltd relates to growth hormone therapy and not diabetes. Mohammed Didi has received payment for advisory work and funding to attend academic meetings and to support a nursing salary from Novo Nordisk Ltd. The work undertaken by Mohammed Didi for Novo Nordisk Ltd relates to growth hormone therapy and not diabetes. Mohammed Didi has received expenses from Merck Serono Ltd (Feltham, UK) to attend educational meetings. Carrol Gamble is a member of the Efficacy and Mechanism Evaluation Programme Board of the National Institute for Health Research. Dyfrig Hughes is a member of the Health Technology Assessment (HTA) Clinical Trials Board (2010–16), the HTA Funding Teleconference (2015–16) and the Pharmaceuticals Panel (2008–12). John W Gregory reports grants from Cardiff University during the conduct of the study, and membership of the HTA Commissioning Board (2010–14). John W Gregory received speaker fees from Pfizer Inc. and Eli Lilly and Company (Basingstoke, UK), fees for attending an advisory board for Eli Lilly and Company and financial assistance to attend annual meetings of the European Society for Paediatric Endocrinology from Sanofi Genzyme (Guildford, UK), Merck Serono Ltd, Ipsen and Novo Nordisk Ltd.

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Blair et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2018 Queen’s Printer and Controller of HMSO

Glycaemic control, intensive insulin therapy and complications of diabetes

In the short term, failure to manage T1D adequately may result in episodes of hypoglycaemia that may result in seizures in the most extreme circumstances or high blood glucose concentrations (hyperglycaemia), which may result in life-threatening episodes of diabetic ketoacidosis (DKA) if left untreated. However, it is the long-term vascular complications of T1D that have the greatest impact on morbidity, mortality and well-being. These can be broadly classified as macrovascular complications (coronary artery disease, peripheral arterial disease and stroke) and microvascular complications (diabetic nephropathy, neuropathy and retinopathy).

The risk of acquiring these complications of T1D is directly related to glycaemic control, duration of T1D, insulin sensitivity and weight. By the time patients enter adult health-care services, many will have lived with T1D for ≥ 10 years, including a period of pubertal growth and development when changes in lifestyle, relationships within families and with peers, normal risk-taking behaviour and physiological insulin resistance make glycaemic control particularly challenging.

In 1993, The Diabetes Control and Complications Research Group3 published conclusive evidence that both the degree and duration of hyperglycaemia are critical determinants of the risk of both microvascular complications and macrovascular complications. The median HbA_1c concentration was lower in patients treated intensively with either MDI or CSII (7.3%) than in patients treated using conventional insulin regimes (9.1%). In patients with established T1D who were observed over a period of 6.5 years, the risk of acquiring retinopathy or nephropathy was reduced by 76% and 39%, respectively, and of having a cardiovascular event was reduced by 41% in those treated with CSII or MDI. Glycaemic control exerted greatest influence on acquisition or progression of complications; however, following correction for the HbA_1c concentration, those patients who were treated with intensive regimes demonstrated a persistent benefit.

Data from historical paediatric cohorts have confirmed that the link between glycaemic control and microvascular complications is established in childhood. 5 Furthermore, there is a strong relationship between improved glycaemic control in children and young people and a fall over time in the prevalence of childhood-onset retinopathy and nephropathy. 6

Glycaemic control in the first year after diagnosis has a long-lasting effect on glycaemic control in later life and, independently of glycaemic control, on the prevalence of complications. Two British, single-centre, retrospective studies, together reporting data from > 300 patients, reported that HbA_1c concentration at 6 months following diagnosis of T1D was an independent predictor of long-term glycaemic control up to 10 years later. 7,8 In a larger Swedish cohort of > 1550 children, HbA_1c concentrations at 3–15 months following diagnosis was related to poorer glycaemic control and increased risk of microalbuminaemia and retinopathy in young adult life. 9 These data highlight the importance of researching interventions that may influence glycaemic control during the critical first year of diagnosis.

Data relating to glycaemic control in childhood and macrovascular complications in adult life are sparse because of the long natural history of the disease. However, markers of an increased risk of cardiovascular complications have been validated and used to identify modifiable risk factors in childhood and adolescence. The SEARCH for Diabetes in Youth (SEARCH) study10 has produced evidence that T1D affects (increases) arterial stiffness, a validated measure of risk for cardiovascular events and mortality in adult life, by late adolescence and early adult life. When data from a mixed cohort of patients with T1D and healthy control participants were analysed using multivariate analyses, other correlates of increased arterial stiffness included adiposity, blood pressure, lipid profile, ethnicity (black and hispanic ethnicities), blood pressure and microalbuminuria. In addition to the long-term vascular complications of poor glycaemic control, patients are subject to a number of health and social disadvantages during childhood. Hospital admission rates are nearly five times higher for children with T1D than for the background population, with the poorest children and those in the youngest age group being most likely to require admission. 11 Frequent episodes of severe hypoglycaemia and sustained hyperglycaemia have been associated with changes on functional brain magnetic resonance imaging, impaired memory and poorer cognitive outcomes. 12,13 Patients with poor glycaemic control may experience poor growth, particularly during puberty,14 and poor glycaemic control has also been associated with an increased risk of childhood depression and use of antidepressants. 15

Prevalence of vascular complications of type 1 diabetes

In the UK, data describing patient demographics, insulin treatment regimes, compliance with NICE guidelines and clinical outcomes are collected in the NPDA. In the 2015/16 NPDA report,1 9.7% of patients had microalbuminuria, an early marker of evolving diabetic nephropathy, and 13.8% had an abnormality on screening for retinopathy. Nearly one-third of patients were hypertensive, 20% had elevated blood cholesterol and 21% of those aged ≥ 12 years were obese. In a quantitative epidemiological systematic review of data from young adults (aged 18–30 years),16 there was some evidence of diabetic retinopathy in 50% of patients, with more severe features presenting in 10% of patients. In the same population, one in six patients had evidence of evolving diabetic nephropathy and half were hypertensive. 16

The data for adult patients for 2014/15 have yet to be reported in the National Diabetes Audit. However, the 2012/13 audit17 reported that, compared with the background population, adults with T1D are 139% more likely to be admitted to hospital with angina, 94% more likely to be admitted to hospital with myocardial infarction, 126% more likely to be admitted to hospital with heart failure, 63% more likely to be admitted to hospital with a stroke, 400% more likely to be admitted to hospital for a major amputation, 817% more likely to be admitted to hospital for a minor amputation and 272% more likely to be admitted to hospital for renal replacement therapy.

There can be little doubt that the burden of T1D complications has a profound impact on the QoL and career prospects of patients. In a study of 91 young adults in their mid-thirties who had lived with T1D for nearly 30 years,18 the mortality rate was 10 times higher than in a control population, owing to diabetic nephropathy and trauma, including suicide. People with T1D were less likely to be employed and more likely to need social welfare. Long-term complications of T1D were the primary predictor of an adverse outcome, with patients with T1D and no complications showing no significant differences from the control population. 18

Treatment of childhood type 1 diabetes and glycaemic control in the NHS

In 2015/16, the NPDA reported data from approximately 28,400 patients aged < 19 years. 1 Intensive insulin regimes were used by 82% of patients: 54% were treated with MDI, having four or more insulin injections a day, and 28% were treated with CSII, an increase from 14% in 2011.

The NPDA also reports trends of improving glycaemic control over time, with the number of patients achieving the target HbA_1c concentration of < 58 mmol/mol, rising from 15.8% in 2012/13 to 27.0% in 2015/16. In the current NICE guideline for the treatment of childhood T1D, the target HbA_1c concentration has been reduced to 6.5% (48 mmol/mol). 19

The Paediatric Diabetes Best Practice Tariff Criteria20 were introduced in the UK in 2012. These allow for a payment of £3189 per patient per year, linked to the provision of core clinical services for paediatric T1D care. For most children’s diabetes services, the additional funding made available by the Paediatric Diabetes Best Practice Tariff Criteria20 enabled the recruitment of more specialist staff and the relative impact of this increase in resources compared with the introduction of CSII therapy on the improvement in glycaemic control reported in the NPDA is unknown.

National databases of children and young people with T1D are maintained in a number of other countries worldwide, enabling clinical outcomes to be compared between countries. In general, data from the UK compare poorly with those from other countries: a publication reported data from 201221 that were held in the NPDA data set, the Prospective Diabetes Follow-up Registry (DPV) from Germany and Austria, and the T1D Exchange database (T1D Exchange) from the USA. The data showed that glycaemic control was the poorest in the NPDA, which also had the lowest rate of CSII usage. In Britain, the disadvantage of poor glycaemic control during childhood appears to persist into adult life. When HbA_1c concentrations were compared between national registries, reporting data from all age groups, from the UK, Germany, Denmark, Latvia and Austria, only Latvian patients had poorer glycaemic control. 22

Evidence for the use of continuous subcutaneous insulin infusion in childhood

Evidence for CSII therapy comes from three areas: single-centre observational studies, national database reports and randomised controlled trials (RCTs).

Acceptability of continuous subcutaneous insulin infusion and multiple daily injections

The success of treatment with either MDI or CSII is likely to be heavily influenced by patients’ and parents’ views on the acceptability of each treatment modality. Studies examining this aspect of treatment are very limited. Data reporting parental stress and QoL in studies comparing clinical outcomes during treatment with MDI and CSII are summarised in Table 1.

Discontinuation rates for CSII are reported to be < 20%, implying a high level of satisfaction with CSII therapy. 50–52 Older children, girls and those with poor glycaemic control are more likely to revert to MDI. Reasons for discontinuing CSII therapy include greater sense of disease, difficulties in doing sports, poorer sense of well-being during CSII therapy, wearing an insulin pump, embarrassment, pain at the site of needle insertion, poor glycaemic control, fear and dislike of frequent blood glucose monitoring. 53–55

Qualitative work undertaken in a study of 19 parents with children aged < 12 years who had changed from MDI to CSII reported the benefits of CSII therapy to include no longer having to administer painful injections, fewer dietary restrictions, improved quality of family life and improved glycaemic control. 56 When patients and parents were interviewed in another study, they also reported improved glycaemic control, in addition to increased lifestyle flexibility and participation in social activities, during CSII therapy. 57 However, parents also reported increased demands on them during CSII therapy, primarily relating to increased blood glucose monitoring. These data give only a limited insight into the acceptability of either treatment. It is likely that patients who participated in these studies changed from MDI to CSII treatment because MDI treatment was unsatisfactory, and studies of unselected cohorts of patients are required to examine acceptability of both treatments.

Costs of treatment of childhood type 1 diabetes and its complications

The cost of T1D to the NHS in the UK is significant: estimates of expenditure range from £1B to £1.8B per year,58,59 and this is expected to rise further, and is projected to account for 2% of total NHS expenditure over the next two decades. 58 The proportion of this cost that is attributable to paediatric T1D services and the budgetary implications of switching patients to CSII is unknown. The number of children and young people aged < 19 years old in the UK with a diagnosis of diabetes is 31,500;60 however, this may be a conservative estimate as not all children aged > 15 years old are managed in paediatric care settings and the true prevalence could be as high as 42,000. The majority of paediatric patients (95.1%) have T1D60 requiring insulin therapy, which is currently administered by CSII in 24% of 10- to 14-year-olds and by MDI in 67%, with the remainder on mixed-methods treatment (three injections per day or fewer). 61 Based on published estimates of MDI and CSII costs62 and the modelled costs of mixed insulins,19 the current annual treatment costs for children and young people with T1D range from £52M to £70M. This would rise to between £79M and £106M if all were converted to using CSII. Additional costs relating to the management of T1D and its complications may be more than three times higher than treatment costs alone,58 suggesting total current annual costs of between £179M and £241M, rising to between £272M and £366M if all patients used CSII.

In 2011, the year that the SubCutaneous Insulin: Pumps or Injections? (SCIPI) study opened to recruitment, it was also estimated that patients with T1D took 830,000 sickness days from work at a cost of £94M, and additional costs associated with T1D during work were > £91M. Premature death accounted for a further 37,000 lost working years, at an estimated cost of £0.6B. 58

It is clear from these observations that T1D represents a significant threat to the well-being and life expectancy of affected patients, a significant cost to society in loss of productivity and economic output and an increasing threat to the NHS as its prevalence continues to rise, particularly in the youngest patients, who will require NHS treatment for the longest period of time. It is critical, therefore, to identify and manage factors that may influence the natural history of the disease and the risk of complications.

Rationale for research

The role of intensive insulin therapy in optimising glycaemic control and thereby reducing the risk of vascular complications of T1D is unquestioned. The optimal way in which to achieve this and the cost-effectiveness of the tools currently available are unknown. A number of observational studies and small RCTs have compared the outcomes of children and young people treated with CSII with those treated with MDI. However, in studies reported to date, there are concerns relating to bias, small patient numbers and short observation periods, and no RCT is directly applicable to the health-care environment in the UK.

Aims and objectives

The SCIPI study is a pragmatic RCT that compares the outcomes of infants, children and young people treated with CSII with those treated with MDI from the time of diagnosis of T1D. The study protocol was developed during a period in which children’s diabetes care in the UK was changing rapidly with the widespread introduction of CSII.

The aim of the SCIPI study was to provide robust clinical data, together with a careful health economics appraisal, to inform the place of CSII therapy in the treatment of individual patients from diagnosis of T1D, and within national diabetes treatment strategies. This study compared CSII with MDI during infancy, childhood and adolescence to identify which treatment facilitates superior glycaemic control and to examine the impact that treatment modalities have on other predictors of vascular complications of T1D, adverse events (AEs) and QoL.

The study was designed with an internal pilot prior to progression to the full study. The objectives of the pilot and full study are detailed in the following sections.

Summary

Type 1 diabetes is a common disease of childhood, associated with complex treatment regimens and lifelong complications that pose a significant burden to individual patients. The cost to society and the NHS is also significant, as patients may not fulfil their potential in the workplace, may be more dependent on state assistance and may need intensive and expensive therapies throughout their lifetime.

There is the potential to modify the disease trajectory of patients by optimising treatment, particularly in children who will live with T1D for many years. Data from national registries suggest that children in the UK have less satisfactory glycaemic control than those in other health economies in which access to health care is also unrestricted. Some researchers have linked improved glycaemic control to higher rates of CSII use; however, the characteristics of those who are treated with CSII are the same as those that favour good glycaemic control, and the relative contribution of these factors is unclear.

Measures taken to address bias in patient populations

Previous RCTs have recruited patients with established T1D who are treated with MDI and randomised them to continue with treatment with MDI or to change to CSII. This recruitment strategy may introduce bias in favour of CSII, as patients who are satisfied with treatment with MDI and who have good glycaemic control are less likely to be approached or to participate in a study than those in whom treatment is less satisfactory. Furthermore, previous experience of treatment success or failure is likely to have an impact on the use of a treatment in the future. In order to address this potential bias, we recruited patients from the time of diagnosis of T1D.

We continued to address the internal pilot study objectives throughout the full study, by maintaining screening logs documenting patients’ demographic data (age, sex, ethnicity and deprivation score derived from their postcode) and the reasons why patients were ineligible, why eligible patients were not approached to participate and why they declined. This enabled us to identify differences in the characteristics of those who consented to participate and those who declined and also to identify major barriers to recruitment.

Measures taken to address potential bias in treatment arms

There are two key areas of concern that were addressed within the SCIPI study design that could otherwise lead to bias between treatment arms.

Measures taken to ensure generalisability of study findings

It was important to generate data that could be used to inform local and national treatment strategies. To achieve this, the SCIPI study needed to recruit and retain participants who were representative of children and young people treated in the NHS. This required that the protocol be deliverable in a wide range of clinics and patient populations, including large clinics in university hospitals and smaller clinics in district general hospitals. To ensure a high retention rate, clinic, participant and parent acceptability needed to be high. The trial design aimed to achieve minimal disruption to normal clinic routines and appointments by drawing on clinical information collected routinely, ensuring that study visits coincided with the times of routine clinic appointments and including minimal additional tasks for patients and their carers.

Study design

The SCIPI study was designed as a pragmatic, open-label two-arm multicentre RCT comparing use of CSII with use of MDI in children and young people aged 7 months to 15 years who were newly diagnosed with T1D. The study included an internal pilot with a sample size of 30 participants to estimate the rate of consent to randomisation and to define the characteristics of recruited participants.

Progression from the internal pilot to the full study was based on the following criteria:

1. At least 50% of patients who are eligible and are invited to participate in the pilot study are successfully recruited.

2. Demographic characteristics that are significantly associated with glycaemic control (age, ethnicity, sex, deprivation score) are not significantly different in the group of patients who are recruited compared with those who decline.

The trial protocol has been published previously. 64 A schematic representation of the study design is given in Figure 1.

FIGURE 1.

Study design. HUI2, Health Utilities Index Mark 2; PedsQL, Pediatric Quality of Life Inventory.

Recruitment procedure

At the time that the SCIPI study opened, children’s diabetes teams were notified within 72 hours of presentation of all infants, children and young people who were newly diagnosed with T1D. This possible delay in the identification of potential participants was addressed when the Paediatric Diabetes Best Practice Tariff Criteria20 were introduced in 2011/12, shortly after the study opened. It is a requirement for centres receiving this additional payment that newly diagnosed patients are discussed with a senior member of the children’s diabetes team within 24 hours of presentation, and that all new patients are seen by a member of the children’s diabetes team on the next working day. All potentially eligible participants were therefore identified promptly.

Treatment group allocation

Education

At entry to the study, all participants completed a structured educational programme delivered to participants and their families in accordance with the standards of the International Society for Pediatric and Adolescent Diabetes. 63

Participants and their families were educated in:

• type 1 diabetes

• the use and administration of insulin

• hyperglycaemia and correction doses

• hypoglycaemia symptoms and treatment

• exercise

• sick day rules

• carbohydrate counting

• the benefits of maintaining optimal glycaemic control for long-term health

• blood glucose monitoring.

All the participants were trained in the use of the MDI regimen and the F. Hoffman-La Roche AG Expert glucometer; participants undergoing CSII treatment were also trained in the use of CSII pumps.

Additional diabetic education was organised to suit individual needs of the participant and family. The dietitian met the participant and their family to assess their diet and educate them in carbohydrate counting. All contact was recorded within the participant follow-up.

To support recruiting sites in the initiation of CSII so soon after diagnosis of T1D, treatment guidelines and written patient information were available from the lead site. These resources were used at the discretion of the recruiting sites.

Baseline assessment

At entry to the study, we collected the following routine biochemical data recorded in medical case notes:

• blood pH at diagnosis

• blood glucose at diagnosis

• presentation with DKA

• glycosylated haemoglobin

• thyroid function tests

• tissue transglutaminase antibody titres, or other screening test for coeliac disease

• titres of anti-glutamic acid decarboxylase and anti-islet cell antibodies.

These baseline tests were not required as part of the SCIPI study protocol, but results were collected if available.

Baseline measurements of height and weight were recorded on the day that the participants commenced randomised treatment to allow for rehydration in participants who were dehydrated at diagnosis of T1D. The Health Utilities Index (HUI) questionnaire was completed and information was collected relating to prescribed concomitant medications. The Health Utilities Index Mark 2 (HUI2) was used for the base-case analysis and the Health Utilities Index Mark 3 (HUI3) (based on a Canadian tariff) was used in the sensitivity analysis.

Follow-up

Study visits were timed to coincide with the times of routine clinic appointments at 3, 6, 9 and 12 months from the date of randomisation. A window of 15 days either side of the appointment time was permitted to allow for holidays, clinic availability and other commitments. A schematic representation of follow-up assessments is given in Table 3. At each visit, permanent and temporary changes in the delivery of insulin were recorded under the heading ‘Review of insulin use’.

Measures

Sample size

A difference in HbA_1c concentration of 0.5% is widely recognised as the threshold used by the US Food and Drug Administration and the pharmaceutical industry to determine effectiveness of any new oral hypoglycaemic agents. A meta-analysis of 20 studies comparing CSII with MDI detected an improvement of 0.61% in adults, suggesting that, in addition to this estimate being the minimum clinically important, it is also a realistic difference to detect. 81–100 To achieve 80% power, a sample size of 143 in each group was required to detect a difference in means of 0.50, assuming that the common SD is 1.50 using a two-group t-test with a 0.05 two-sided significance level. Allowing for 10% loss to follow-up increased the sample size to a total of 316 participants (158 per group). The estimate used for the SD in the sample size calculation was taken from an unpublished audit at Alder Hey Children’s NHS Foundation Trust (Joanne Blair, Mohammed Didi, Princy P, Atrayee Ghatak, Alder Hey Children’s NHS Foundation Trust, 2009, personal communication) based on children matching the inclusion criteria for this proposed study.

Statistical methods

The analysis and reporting of the SCIPI study was undertaken in accordance with the Consolidated Standards of Reporting Trials (CONSORT)101 and the International Conference on Harmonisation E9 guidelines. 102 The main features of the statistical analysis plan are included here with a full and detailed statistical analysis plan provided as a separate document [URL: www.journalslibrary.nihr.ac.uk/programmes/hta/081439/# (accessed 26 July 2018)]. All statistical analyses were undertaken using SAS^® software (version 9.2; SAS Institute Inc., Cary, NC, USA). SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ^® indicates US registration.

A two-sided p-value of ≤ 0.05 was used to declare statistical significance for all analyses. Similarly, all CIs were calculated at the 95% level. No adjustment for multiplicity was made to adjust the type 1 error rate for secondary outcomes. Relevant results from other studies already reported in the literature were taken into account in the interpretation of results.

The primary analysis followed the intention-to-treat (ITT) principle as far as practically possible; a secondary analysis using the per-protocol approach was conducted for the primary outcome. The purpose of the per-protocol approach is to consider the robustness of the conclusions reached from the analysis using the ITT principle, which includes protocol deviations.

Glycosylated haemoglobin, a continuous outcome, was compared between the trial groups using mixed-model regression with 12-month HbA_1c as the dependent variable, treatment group as an explanatory factor and the randomisation stratification variables (age group and centre) as covariates; centre was fitted as a random effect. The mean and SD of HbA_1c concentrations were reported for each age group and treatment group. The mean difference in HbA_1c concentrations and 95% CI between treatment groups was given as the estimated age group- and centre-adjusted treatment effect calculated by the fitted mixed-model regression. Central measures of HbA_1c concentration were used in preference to local measurements; local measurements were used if central ones were not available.

Secondary continuous outcomes were analysed as per the primary outcome methods. For binary outcomes, the number and percentage of participants with the outcome are reported overall and for each treatment group. The difference between groups is tested using the chi-squared test. Relative risks with 95% CI are presented.

The safety analysis data set contains all participants who were randomised and received at least one dose of insulin via the randomised treatment. Participants’ AEs/serious adverse events (SAEs) were included in the treatment group the participant was actually receiving at the time of AE/SAE onset to take into account any participants who permanently changed their mode of insulin delivery at any point during the trial. The number of related AEs occurring and the number and percentage of participants involved were reported by treatment arm. The incidence rates of total numbers of AEs/SAEs were calculated for each treatment group in person-days using the incidence density ratio (IDR). The IDR is the number of patients with at least one new AE per population at risk in a given time period. The denominator is the sum of the person-time in years for each treatment group (accounting for treatment switches) of the at-risk population.

The statistical analysis of the data follows the standard operating procedures of the Clinical Trials Research Centre, which requires independent programming of the primary outcome and safety analyses by an independent statistician.

Study oversight and role of funders

The Trial Management Group (TMG), comprising the chief investigator, other lead investigators (clinical and non-clinical), members of the MCRN CTU and three parent contributors, was responsible for the day-to-day running and management of the trial. The membership of the oversight committees was suggested by members of the TMG to the trial funders and appointed by the funders with their constitution following funder requirements.

The Trial Steering Committee (TSC) consisted of an independent chairperson, Dr Peter E Clayton, two independent experts in the fields of diabetes and endocrinology, Dr Christine P Burren and Dr Ian Craigie, an expert in medical statistics, Professor Gordon D Murray, and a parent contributor, Mrs Christina McRoe. The role of the TSC was as the executive decision-making committee considering the recommendations of the Independent Data and Safety Monitoring Committee (IDSMC). Monitoring reports viewed by the TSC were not split by treatment group.

The IDSMC consisted of an independent chairperson, Professor Stephen Greene, plus two independent members, Professor John Wilding, an expert in the field of endocrinology, and Dr Arne Ring, an expert in medical statistics. The IDSMC was responsible for reviewing and assessing recruitment, interim monitoring of safety and effectiveness, trial conduct and external data. The IDSMC provided recommendations to the TSC concerning the continuation of the trial and viewed accumulating data split by treatment group.

Recruitment

The study opened to recruitment on 16 May 2011 and was closed on 30 January 2017.

At the outset of the study, we planned to recruit 316 patients within 30 months across eight study sites. The recruitment rate had been informed by the DECIDE trial68 and the study sites were selected to build on the experience and expertise acquired by local researchers during the DECIDE trial.

All study sites had to secure funds to meet the excess treatment costs of the study protocol. This process incurred some considerable delays, and in some centres was not possible. Recruitment rates were also slower than expected and, consequently, the recruitment curve was revised in month 18 to allow for an additional 12 months of recruitment and again in month 40 to allow a further 6 months. The number of study sites was increased from 8 to 15. The recruitment graph showing the initial, and revised, predictions and the observed recruitment numbers is displayed in Appendix 1.

To increase recruitment, key protocol amendments were made (protocol version 3.0 to 4.0) that increased the permitted time window for recruitment from diagnosis to 14 days and softened eligibility criteria to place emphasis on ability to comply with treatment regimens rather than complete study questionnaires and to support inclusion of children with parents with T1D but to maintain exclusion of siblings with T1D.

To optimise consent rates, screening log data were used to inform the recruitment strategy. Screening data indicated that patients who consented to participate were approached sooner after diagnosis than those who declined. In the light of these data, clinical teams and research nurses (RNs) were encouraged to share information about the study as soon as possible following diagnosis.

In total, 976 patients were diagnosed with T1D in the 15 study centres in England and Wales during the 48-month recruitment period, of whom 98 (10%) did not meet the study eligibility criteria. One hundred and eighty-nine patients (22%) were not approached about participation. Of these, 115 (61%) were considered to be unsuitable by treating clinicians for reasons other than the study exclusion criteria, 52 patients (28%) were not approached because staffing levels at the site were too low at the time of diagnosis to deliver the study protocol and no reason was recorded for 22 (12%). Of the remaining 689 patients, 294 (42.7%) consented to participate in the study; however, one randomised participant withdrew consent for their data to be used immediately following randomisation, leaving 293 randomised participants.

Patients and families were invited to share their reasons for declining to participate in the study. The main reason given was patient preference for one of the trial arms. Strong patient preference had been expected at the design stage, and of the 395 eligible participants who declined consent, 36 (9%) cited a preference for CSII therapy and 259 (66%) cited a preference for MDI.

In response to this strong patient preference and to support an informed decision, we made a short film in which four SCIPI study participants (two randomised to CSII and two randomised to MDI) and their parents shared their experiences of diagnosis and living with T1D and the treatment in each of the study arms. Although the film was generally well received, we did not observe any increase in recruitment rates following its introduction and the issue of patient preference persisted throughout the trial.

A CONSORT flow diagram illustrating the pathway of patients from diagnosis to consent and randomisation and through the study protocol is given in Figure 2.

FIGURE 2.

The CONSORT flow chart.

The purpose of the SCIPI study was to generate evidence that could be applied to the population of patients treated in children’s diabetes services in the NHS. For this reason, it was important to demonstrate that patient characteristics, in particular those known to be associated with glycaemic control, did not differ between those who consented and those who declined to participate. The baseline characteristics of patients who were invited to participate in the study are given in Appendix 1, Table 28. There was no difference in age, sex or ethnicity between the group of patients who consented and the group of patients who declined. The median deprivation score for those who declined because they had a strong preference for CSII was 27.7 (range 16.6–63.9), compared with 17.0 (range 1.62–77.23) for those who consented and 17.96 (range 1.18–74.35) for those who stated a strong preference for MDI.

Internal pilot

The internal pilot was planned after 30 patients had been recruited. Of the 89 eligible patients approached for consent, 30 provided consent and 59 declined (33.7% consent rate, 95% CI 23.9% to 43.5% consent rate). The first patient was randomised on 31 May 2011, and 30 patients were randomised by 3 July 2012. A review of the consent rates by recruiting centre identified lower consent rates in one centre that started patients on three daily injections at diagnosis, rather than MDI. Excluding this site, the consent rate increased (26/60 patients; 43.3%, 95% CI 30.8% to 55.9%). Screening data comparing characteristics of eligible consenting participants did not differ from those declining for age, sex, ethnicity or deprivation score. A review of the SD used in the sample size calculation to the accrued data indicated that the study should continue to the planned sample size. The recommendation of the IDSMC was that the SCIPI study should progress to the full study.

Baseline comparability

There was no difference in age, sex, ethnicity or deprivation score between treatment arms (Table 4). Baseline data also showed that auxological and biochemical characteristics did not differ between treatment groups at baseline (Table 5).

The measurement of autoantibodies was determined by local protocols, with some centres testing only patients in whom the diagnosis of T1D was felt to be uncertain. Eleven patients randomised to the MDI arm tested positive for coeliac disease on antibody testing, compared with four randomised to the CSII arm.

Retention and adherence

Figure 3 summarises the time point and number of patients who made a permanent change to their randomly allocated insulin delivery method, the time point and number of patients who decided to end their follow-up in the SCIPI study and the number of patients who attended each scheduled follow-up. In total, only six participants withdrew from the SCIPI study prior to the 12-month follow-up, and all remaining 287 patients (97.0%) attended the 12-month follow-up visit. See Figure 2 for a summary of the reasons for withdrawals.

FIGURE 3.

Retention and adherence. a, Permanent change of insulin delivery from randomised treatment but continuing follow-up; b, patient changed from MDI to injections ter die sumendus (TDS) regime.

Primary outcome: glycosylated haemoglobin measured at 12 months

Data on HbA_1c concentrations were available for 97% of participants (CSII, n = 143; MDI, n = 142).

The results for the primary outcome for the primary ITT analysis and the per-protocol analysis are provided in Table 9. The per-protocol analysis population excludes participants who had a major protocol deviation (see Table 8) or three or more minor deviations.

The HbA_1c measurement was marginally higher in the CSII group than in the MDI group (mean difference 2.4 mmol/mol; 95% CI –0.4 to 5.3 mmol/mol), although the observed difference was not statistically significant. This result was consistent when compared with the per-protocol analysis (see Tables 4 and 5 for comparison of baseline characteristics of all randomised participants and Appendix 2, Table 30, for baseline characteristics of the per-protocol analysis population). A summary of the data sources (local or central laboratory) used in the primary ITT analysis and the per-protocol analysis is provided in Table 10.

Exploratory analyses including HbA_1c concentrations measured at baseline as a continuous explanatory variable in the age-adjusted model did not alter the SCIPI study conclusions for CSII compared with MDI at 12 months [least square mean difference between treatment groups (CSII – MDI) 2.9; 95% CI –0.02 to 5.9], but did suggest the importance of early baseline values for 12-month measurements (HbA_1c concentration baseline coefficient estimate 0.07; standard error 0.03; 95% CI 0.01 to 0.13).

An additional exploratory analysis considered the impact of the deprivation score by including it in the age strata-adjusted model and found that it did not alter conclusions [least square mean difference (CSII – MDI) 2.2; 95% CI –0.7 to 5.0]. The model suggested that higher deprivation score was associated with higher HbA_1c concentrations at 12 months (0.03, standard error 0.04, 95% CI –0.05 to 0.12).

Mean profile plots are provided for each age stratum in Figure 4. The results from including an age stratum by treatment group interaction with age as a main effect in the model are provided in Table 11. The least mean square estimates are 61.5 (95% CI 59.1 to 63.9) for CSII, 57.9 (95% CI 55.5 to 60.3) for MDI and 3.6 (95% CI 0.6 to 6.6) for the difference (CSII – MDI).

FIGURE 4.

Glycosylated haemoglobin (mmol/mol) mean profile plots by age stratum. (a) Participants aged < 5 years; (b) participants aged 5–11 years; and (c) participants aged ≥ 12 years. Not a prespecified analysis in the statistical analysis plan.

Figure 5 displays the consistency of treatment effect across the SCIPI study centres. The consistency of treatment effect across sites is evident by the overlapping CIs. There is no statistical evidence for differences between sites from the chi-squared test for heterogeneity (p = 0.51) and I² (0%). Figure 6 displays the treatment effect in 6-month periods to consider potential presence of a learning curve from the first randomisations. The period of 0–6 months includes the first 6 months of randomisations within each centre. The graph does not suggest a pattern of reducing treatment effect by time.

FIGURE 5.

Forest plot of HbA_1c concentrations at 12 months by site. IV, inverse variance.

FIGURE 6.

Forest plot to consider treatment effect learning curve. Not prespecified analysis in the statistical analysis plan. IV, inverse variance.

Secondary outcomes

The target HbA_1c value that was recommended by NICE19 at the start of the SCIPI study was < 58 mmol/mol. In August 2015, this was revised to be < 48 mmol/mol. To reflect current practice, the secondary outcome that had previously been < 58 mmol/mol was changed in a protocol amendment (version 7.0). However, as the majority of the SCIPI study participants would have been randomised and followed up with diabetes teams aiming to attain HbA_1c values of < 58 mmol/mol, the results for both are presented in Table 12. Greater proportions of participants on the MDI arm attained values below the target thresholds; however, results are not statistically significant.

The incidences of severe hypoglycaemia and DKA were low, with only eight and two participants experiencing these events, respectively. Although participants on CSII had a greater event rate, for both events, the wide CIs reflect the low number of event rates and, hence, a high level of uncertainty.

A higher proportion of participants randomised to MDI were in partial remission at 12 months. However, levels of missing data were high for this outcome because of the lack of availability of the daily insulin dose data required to calculate its occurrence. The formula used to calculate partial remission is available in the statistical analysis plan, which is available on the National Institute for Health Research website (www.journalslibrary.nihr.ac.uk/programmes/hta/081439/#/; accessed 13 February 2018).

There was no significant difference in change in BMI SDS or height SDS between those randomised to CSII and those randomised to MDI (Table 13). Insulin requirements were significantly lower for those randomised to MDI than for those randomised to CSII (see Table 13). Similarly, with partial remission, the low availability of dose data increased levels of missing data. Owing to the availability of CSII data, levels of missing data are higher in the MDI arm.

Quality of life

Tables 14 and 15 present results for QoL as measured by the PedsQL instruments for parents and children, respectively, stratified by instrument age category. Results for each domain are available in Appendix 2, Tables 31 and 32.

In parent-reported QoL, there was a tendency for parents with children randomised to CSII to report higher QoL scores than those on MDI. This was consistent across instrument age strata, the 6- and 12-month follow-up time points and the child-reported scores. Estimates of the least mean squares difference between treatment groups (CSII – MDI) were 4.1 (95% CI 0.6 to 7.6; p = 0.02) for parent-reported and 3.1 (95% CI –0.6 to 6.8; p = 0.1) for child-reported QoL. It should be noted, however, that child-reported QoL starts from age 5 years but the parent-reported QoL starts from age 2 years and that, in general, the size of the differences and CI widths are similar.

Safety analysis

Table 12 reports on incidences of hypoglycaemia and DKA in the ITT analysis population. In this section, the data set contains all participants who were randomised and received at least one dose of trial medication. For AEs, the method of insulin delivery that the participants were receiving at the time of AE onset was used, rather that the allocated treatment, to take into account that participants could temporarily or permanently switch treatment. The total number of events experienced and the number of participants experiencing at least one event are provided along with the IDR (the number of patients with at least one new AE per population at risk in a given time period):

• The incidence of related AEs by treatment group was 54 in 36 patients who were on CSII at the time of the AE. IDR was 25.0 patients with at least one event per 100 person-years and 17 AEs in 16 patients who were on MDI at the time of the AE. IDR was 10.5 patients with at least one event per 100 person-years.

• The incidence of related SAEs by treatment group was 14 SAEs in nine patients who were on CSII at the time of the SAE. IDR was 6.2 patients with at least one event per 100 person-years and eight SAEs in eight patients who were on MDI at the time of the SAE. IDR was 5.3 patients with at least one event per 100 person-years.

Tables 16 and 17 provide the summary of related AEs and SAEs, respectively.

Introduction

Aim

The aim of the economic evaluation of the SCIPI study was to estimate the cost–utility, based on an assessment of the incremental costs per QALY gained, of treating paediatric T1D patients with CSII versus MDI. The costing perspective was that of the NHS in England. Within-trial QALYs were derived from patients’, parents’ or guardians’ responses to the HUI Quetionnaire using the HU12 algorithm. Costs were derived by measuring health-care resource use associated with both CSII and MDI, including (1) purchase of pumps and pen devices, (2) consumables, (3) management of AEs, (4) procedures, (5) insulin, (6) hospitalisations and (7) contact with health-care professionals including GPs, school nurses, as an outpatient and at the accident and emergency (A&E) department. Differences between intervention groups in the primary outcome would indicate the need for a lifetime extrapolation using the CDM, otherwise a 12-month time horizon would be adopted.

Methods

Results

Health utilities and quality-adjusted life-years

The HUI questionnaires were not completed for all patients. Thirteen of the 14 patients for whom there were no HUI data were aged < 3 years at randomisation and so were excluded from the analysis. Baseline utilities were missing for 23 patients in the CSII group and 14 patients in the MDI group.

For patients with complete baseline utility assessment, mean values were 0.870 (95% CI 0.847 to 0.892, n = 115) for the CSII group, compared with 0.888 (95% CI 0.866 to 0.909, n = 127) for the MDI group. Two hundred and fourteen patients (CSII group, n = 109; MDI group, n = 105) were included in the base-case analysis. Changes in utilities and each of the HUI domains (vision, hearing, speech, ambulation, dexterity, emotion, cognition and pain) over the 12-month period are presented in Figures 7 and 8, respectively. Both intervention groups provided improvements in patients’ happiness (emotion) and how they perceived pain perception. However, there were no significant differences in QALYs between intervention groups. For the CSII group, these were 0.910 QALYs (95% CI 0.892 to 0.927 QALYs), and for the MDI group these were 0.916 QALYs (95% CI 0.899 to 0.933 QALYs), which was a difference of –0.006 QALYs (95% CI –0.031 to 0.018 QALYs).

FIGURE 7.

Mean utility values (95% CI) over 12 months for the base-case analysis.

FIGURE 8.

Distribution of responses to HUI domain, by intervention group, over time. (a) Vision; (b) hearing; (c) speech; (d) ambulation; (e) dexterity; (f) emotion; (g) cognition; and (h) pain.

Uncertainty analysis

The cost-effectiveness plane, which presents the difference in costs on the vertical axis and difference in effects on the horizontal axis, is shown in Figure 9. Sixty-nine per cent of iterations were in the north-west quadrant, reflecting the probability of CSII being dominated by MDI. The corresponding cost-effectiveness acceptability curve (Figure 10) indicates that CSII is not cost-effective, even at very high thresholds, consistent with the certainty of there being a large cost difference with no improvement in QALYs.

FIGURE 9.

Cost-effectiveness plane presenting the joint uncertainty in incremental costs and QALYs.

FIGURE 10.

Cost-effectiveness acceptability curve indicating the probability of CSII being cost-effective for different thresholds of cost-effectiveness. (a) Proportion 0–1; and (b) proportion 0–0.01.

Summary of findings

The results of the SCIPI study clearly demonstrate that, across the range of ages from 7 months to 15 years, treatment with CSII is not superior to treatment with MDI during the first year of treatment for T1D and is not a cost-effective treatment. At 12 months following diagnosis, HbA_1c concentrations differed by only 2.4 mmol/mol (p = 0.1), being lower in participants treated with MDI and with our study having been powered to detect a difference of > 0.5%, approximately 5 mmol/mol. The percentage of participants whose HbA_1c concentration was within the target range (< 48 mmol/mol) at 12 months differed between the treatment arms by 5% (CSII, 15.4%; MDI, 20.4%) and was also not significantly different (p = 0.28).

The proportion of patients in the CSII group who experienced AEs (28%) was nearly three times that in the MDI group (10%). During the course of the study, there were 11 severe AEs (nine episodes of severe hypoglycaemia and two of DKA), of which nine occurred in patients treated with CSII.

Patient-reported disease-related QoL did not differ between treatment arms, although parent-reported measures of QoL showed higher scores for patients who were treated with CSII, indicating a superior QoL. However, it should be noted that child-reported QoL starts from age 5 years, whereas the parent-reported QoL starts from age 2 years, and that, in general, the size of the differences and CI widths are similar between child- and parent-reported measures.

Patients randomised to CSII had more than twice as many A&E department visits and inpatient stays relating to the management of T1D as those in the MDI group. Over the 12-month period, health-care professionals had, on average, 4.3 (95% CI 0.6 to 8.0) more contacts (texts, e-mails and telephone calls) with participants in the CSII group than with those in the MDI group. The mean total costs were £1863 (95% CI £1620 to £2137) higher in the CSII group than in the MDI group, with the majority of this difference (£1177) attributable to the additional cost of consumables and device (annualised cost of £600 for CSII vs. £80 for pen devices). There were no significant differences in QALYs between the CSII and MDI groups (–0.006 QALYs, 95% CI –0.031 to 0.018 QALYs).

In summary, the SCIPI study achieved excellent completeness of primary outcome data for assessment of efficacy of CSII compared with MDI. There were no differences in key baseline characteristics between groups at randomisation. There was no difference in efficacy between CSII and MDI in any of the primary ITT analyses. In addition, there was no difference in efficacy between CSII and MDI in the per-protocol ITT analyses. The economic evaluation indicated that CSII does not represent a cost-effective option for the management of T1D in patients representative of the SCIPI study population. There was also no evidence of a difference in QALY gain over the 12-month period of observation in participants treated with CSII compared with those receiving MDI, and the cost differential was such that, even for willingness-to-pay values far in excess of the NICE £30,000 per QALY threshold for cost-effectiveness, CSII remained dominated by MDI.

Patient recruitment and retention

The large and diverse population of patients recruited to the SCIPI study, coupled with a high retention rate (97%), enabled us to overcome elements of bias inherent in previous RCTs and observational studies. The characteristics of study participants show that the treatment arms were well balanced for factors known to influence glycaemic control, including deprivation score, sex and ethnicity. Screening log data show that these characteristics were also well balanced between the groups of patients who consented to participate in the study and those who declined. The distribution of ages in the study population reflects the known demographic of childhood T1D, with approximately 50% of participants being in the age range of 5–12 years.

Of those who declined to participate, most (66%) cited a strong preference to continue treatment with MDI, although a smaller number (9%) cited a strong preference for CSII. We noted that the small number of patients and parents who declined to participate in the study, stating a strong preference for CSII (n = 36), had a higher deprivation index (were more deprived) than those who declined stating a strong preference for MDI (n = 259) and those who participated in the study. National database studies37–39 report an association with CSII use and deprivation, with the most deprived patients being least likely to be treated with CSII. Our data suggest that this association is established sometime later after the diagnosis of T1D.

Recruitment rates showed marked variability between study sites, ranging from 16% to 82% of eligible patients. Both the highest- and lowest-recruiting centres had participated in the DECIDE trial68 and so were experienced in the recruitment of paediatric patients to research studies at the time of diagnosis of T1D. In the lowest-recruiting centre, intensive insulin therapy was introduced later in the first year of treatment and patients were treated with four or fewer injections per day at the time of diagnosis. A requirement to either increase the number of injections or to learn a new method of insulin delivery is likely to have been least attractive to this population.

We recognised that patients who were given information about the study closest to the time of diagnosis were more likely to consent than those who were approached later. This may be because a preference for MDI had yet to be established or may reflect qualities of the research staff that increased the likelihood of patients consenting to participate in the study.

We speculate that the strong preference that patients stated for MDI at the time they were invited to participate in the study is likely to be temporary. It would have been ideal to undertake qualitative work to understand more about the reasons for this preference and how it may change over time. A higher recruitment rate may have been achieved had the window for recruitment been increased to 1 month from diagnosis.

These observations underscore the difficulty in predicting recruitment rates to clinical trials. The target recruitment rate of 50% was set, in part, to ensure that a highly selected cohort of patients was not recruited. Despite our lower recruitment rate, we have been able to demonstrate that the characteristics of those who consented to participate did not differ significantly from those who declined, and we are confident that we have studied a population of patients representative of the background population of paediatric patients treated for T1D in the NHS setting.

The non-participation of patients with a strong preference for one treatment arm may limit the generalisability of our data. It would be interesting to learn more about the reasons behind this strong preference early in the diagnosis of T1D and how this may change over time. A qualitative component within the SCIPI study would have been beneficial.

By randomly allocating treatment to either MDI or CSII, we have overcome the bias in observational studies, in which treatment selection by patients, carers and health-care professionals results in over-representation of favourable patient characteristics in those treated with CSII and under-representation of these characteristics in those treated with MDI.

Previous RCTs of CSII and MDI have recruited patients with established T1D already being treated with MDI, and have randomised them to continue with MDI treatment or to change to CSII treatment. This recruitment strategy may introduce bias in favour of CSII, as patients who are satisfied with treatment with MDI and who have good glycaemic control are less likely to be approached or to participate in the study than those in whom treatment is less satisfactory. Furthermore, previous experience of treatment success or failure with MDI may influence treatment outcomes: if a patient dissatisfied with MDI treatment consents to participate in a RCT, it is possible that randomisation to CSII will result in renewed commitment to the management of T1D; however, randomisation to continue unsatisfactory treatment with MDI may result in further disengagement. Another factor to consider is the extent to which any studies reporting benefits of CSII over MDI in established patients reflect the effects of the intensive education and increased contact with health-care professionals required before starting CSII rather than inherent benefits arising directly from CSII. These possible effects on outcomes were balanced in our study design by the selection of newly diagnosed patients, all of whom would have required a similar intensity of education. We elected to recruit patients at diagnosis of T1D to overcome this element of bias. However, it may be that the generalisability of our findings beyond the first year of T1D is therefore limited.

Fifty-three patients moved between treatment arms: 30 changed from MDI to CSII, 22 from CSII to MDI and one from MDI to ‘injections TDS regime’. In one study site, four out of five patients who were randomised to MDI were changed to CSII and none from CSII to MDI, and, in another site, 9 out of 16 patients who were randomised to MDI were changed to CSII and 3 out of 14 changed from CSII to MDI. It is likely that there is a degree of bias towards CSII in these centres and this may account, in part, for the minor imbalance in movement between treatment arms. We also observed an imbalance in movement between treatment arms in the youngest age group, favouring a switch from MDI to CSII. This may also reflect a bias towards treatment with CSII in this age group, perhaps because of the difficulties of administering frequent injections in children in an age group who eat frequently and unpredictably throughout the day. The administration of very small insulin doses in the youngest children can also be difficult using MDI and this can be accommodated more easily using CSII.

When we examined the reasons for changing between treatment arms, we found that HbA_1c concentration was above the target range in 20 out of 22 patients (90.9%) at the time of changing from CSII to MDI and in 16 out of 30 patients (53%) at the time of changing from MDI to CSII. Severe or frequent hypoglycaemia was not reported as a reason for changing treatment arm. Our data suggest that, following randomisation, patient and parent preference was the strongest determinant of insulin delivery device.

Effectiveness of continuous subcutaneous insulin infusion and multiple daily injections therapy

Implications for diabetes treatment strategies

The purpose of the SCIPI study was to generate objective data to inform national paediatric treatment strategies and individual patient care. We speculated that, in the absence of benefit of CSII therapy, resources may be better invested in services that are likely to benefit a wider range of patients.

In the NPDA of 2015/16,1 2800 children aged < 15 years were diagnosed with T1D. Given that 30% of patients are currently treated with CSII, our data suggest that the additional cost to the NHS of treating newly diagnosed patients with CSII in preference to MDI would reach approximately £1.5M per year, with no objective evidence of benefit, at least in the year following diagnosis.

The data from the SCIPI study should reassure families that there is no evidence that MDI therapy is inferior to CSII during the first year of T1D treatment. We have provided new evidence to clinicians that the driver for the adoption of CSII therapy at diagnosis of T1D should not be improved glycaemic control or AEs; indeed, we would encourage consideration of the increased prevalence of AEs in those treated with CSII.

The social benefits of CSII have not been quantified in this study, and perhaps the balance between cost and benefits in this domain is an issue for wider discussion between funders, health-care professionals, patients and families.

Future research

The findings of the SCIPI study may be applicable only during the first year following diagnosis of T1D. On completion of the study protocol, participants and their families were invited to consent for the collection of routine clinical data for a further 9 years. We therefore have the opportunity to learn more about the longer-term trajectories of this valuable and unique cohort of patients and whether or not outcomes diverge between treatment groups over time.

It is important to recognise that many patient advocacy groups and health-care professionals have a strong belief in the benefit of CSII. The measures of QoL used in the SCIPI study may not have been sensitive enough to detect the QoL benefits reported by the advocates of CSII, and we have not measured generic QoL or the impact of either treatment on social function across childhood and into adolescence. More detailed investigation in these domains may contribute to the critical analysis of the place of CSII in national diabetes strategies.

The need to examine how resources are distributed has become increasingly evident as relationships between socioeconomic status, CSII therapy and glycaemic control are reported. Observational cohort studies from across the world report superior glycaemic control and increased CSII use in children from more affluent families, and their parents have lower unemployment rates and higher educational levels than those children treated with MDI, whereas socioeconomic deprivation and ethnic minority status are associated with poor glycaemic control, independent of CSII use. 38,39,156,157 Unless efforts are made to correct this disparity, health-care resources will continue to be drawn away from the most vulnerable patients to support CSII therapy, with little evidence of benefit. We propose that research should now focus on interventions that may help to address the disparity in outcomes between the most affluent and disadvantaged children.

There is currently no validated, structured education programme for paediatric patients who are newly diagnosed with T1D, and previous studies of structured education have not demonstrated a beneficial effect on glycaemic control. 158 It may be that education to establish good management habits at the time of diagnosis will be more effective than in established patients, in whom poor management habits are likely to be entrained and instigating change may be difficult. The benefits of structured, validated education at diagnosis of T1D deserve further evaluation.

In this study, we did not record whether or not patients used CSII to its full potential. It is possible that enhanced CSII use would have resulted in improved glycaemic control and that this may be achievable through further education, training and support in the community. There may be a place for researching these elements of CSII use. However, such a package of care is likely to require increased investment, and this additional cost needs to be considered in the light of the health economic data reported in the SCIPI study. Very significant improvements in glycaemic control would be necessary for CSII to be seen as cost-effective. Should such a project be undertaken, it would be essential to also examine the effect of enhanced education and support in the community on glycaemic control in children treated with MDI.

Contributions of authors

Joanne Blair (Chief Investigator and Consultant Paediatric Endocrinologist) developed the study protocol in collaboration with co-investigators. She oversaw the delivery of the study, prepared study update reports, oversaw clinical aspects of the statistical analysis plan and clinical interpretation of study data and led the preparation of the final report. She was Chairperson of the Trial Management Committee.

Carrol Gamble (Senior Statistician) contributed to the study design, protocol development, and data capture methods and led the CTU involvement including all monitoring activities, the development of the statistical analysis plan and the study statistical analyses. She contributed to the preparation of the final report (drafting, reviewing and editing).

Andrew McKay (Study Statistician) undertook the final statistical analyses under the supervision of Carrol Gamble, prepared data for reports throughout the study, prepared data tables and figures for the final report. He was a member of the Trial Management Committee.

Dyfrig Hughes (Lead Health Economist) contributed to protocol development, oversaw the health economic aspects of the study and the final analyses, contributed to the final report (drafting, reviewing and editing). He was a member of the Trial Management Committee.

Colin Ridyard (Health Economist) undertook health economics analysis, managed all aspects of health economics elements of the study, contributed to the final report (drafting, reviewing and editing). He was a member of Trial Management Committee.

Matthew Peak (Co-Investigator and Director of Research, Alder Hey Children’s Hospital) contributed to protocol development, provided guidance on study delivery, supported interpretation of clinical data and contributed to the final report (drafting and reviewing). He was a member of the Trial Management Committee.

John W Gregory (Co-Investigator, Professor in Paediatric Endocrinology and Honorary Consultant) contributed to protocol development, provided guidance on study delivery, supported interpretation of clinical data and contributed to the final report (reviewing and editing). He was a member of the Trial Management Committee.

Mohammed Didi (Co-Investigator and Consultant Paediatric Endocrinologist) contributed to protocol development, provided guidance on study delivery, supported interpretation of clinical data and contributed to the final report (reviewer). He was a member of the Trial Management Committee.

Francesca Annan (Co-Investigator and Paediatric Diabetes Dietician) contributed to study design, prepared educational materials for study sites and contributed to the final report (reviewer).

Emma Bedson (Senior Trials Manager) gave guidance and support on all aspects of governance and study delivery and supported the preparation of progress reports. She contributed to the final report (reviewer) and was a member of the Trial Management Committee.

Keith Thornborough (RN and Paediatric Diabetes Nurse Specialist) recruited patients, provided guidance and expertise to RNs in study sites, led RN teleconferences and face-to-face meetings and supported interpretation of clinical study data. He contributed to the final report (reviewer) and was a member of the Trial Management Committee.

Publications

Blair JC, Peak M, Gregory JW. What is the best way to deliver subcutaneous insulin to infants, children, and young people with type 1 diabetes mellitus? BMJ 2011;343:d5221.

Blair J, Gregory JW, Peak M. Insulin delivery by multiple daily injections or continuous subcutaneous insulin infusion in childhood: addressing the evidence gap. Pract Diabetes 2012;29:47–8.

Blair J, Gregory JW, Hughes D, Ridyard CH, Gamble C, McKay A, et al. Study protocol for a randomised controlled trial of insulin delivery by continuous subcutaneous infusion compared to multiple daily injections. Trials 2015;16:163.

Arch BN, Blair J, McKay A, Gregory JW, Newland P, Gamble C. Measurement of HbA_1c in multicentre diabetes trials – should blood samples be tested locally or sent to a central laboratory: an agreement analysis. Trials 2016;17:517.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

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

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.