Relative effectiveness of insulin pump treatment over multiple daily injections and structured education during flexible intensive insulin treatment for type 1 diabetes: cluster randomised trial (REPOSE)BMJ 2017; 356 doi: https://doi.org/10.1136/bmj.j1285 (Published 30 March 2017) Cite this as: BMJ 2017;356:j1285
All rapid responses
We appreciate the comments and interest of Dr Thabit and colleagues. They highlight limitations of our trial, many of which we commented on in the paper but raise some additional concerns, which for reasons of space we were unable to address. We draw the attention of the authors to the HTA monograph, recently published, which allows us to discuss these issues in more detail.
The authors question the generalisability of our trial on the basis that we excluded patients who wished to use insulin pumps. But as we pointed out in our paper, our study by design meant we should exclude patients who had expressed a wish to use insulin pumps. If we had recruited such individuals then their randomisation to the multiple injection (MDI) arm might have provided misleading outcomes due to disappointment or withdrawal from the trial. We also explained that we excluded patients who met NICE criteria because the case for pumps (CSII) in that group is not in doubt. We have written to NICE to assure them that the results of REPOSE have no implications for the guidance in TA151 . Our aim was to explore whether there was merit in extending indications for insulin pumps to a wider range of patients, including individuals whose HbA1c levels were far from target and who had yet to receive structured education to aid them in undertaking the complex skills of insulin self-management.
As regards the suggestion that training in pump skills might not have been sufficient, we reported in our discussion that the centres participating in the trial were experienced pump centres whose educators were confident that the necessary skills of pump management were provided to pump participants. We are unaware of evidence that the more sophisticated pump features lead to improved glycaemic outcomes although participants were specifically encouraged to seek additional input from educators to guide them in using additional features.
We acknowledge that in terms of the primary outcome the confidence intervals do not exclude a minimally important clinical difference of 0.5% HbA1c although Dr Thabit and colleagues appear to disregard our finding that both structured education and pump arms demonstrated clinically relevant falls in absolute HbA1c of 0.6% as well as falls in rates of severe hypoglycaemia of 50%. Furthermore, our primary outcome, showing an 0.24% difference in HbA1c between the arms at 2 years is very close to the findings of two systematic reviews funded by HTA [3,4] which examined a large number of trials comparing MDI and pump therapy. Even if the difference in HbA1c had reached a clinically significant difference, cost-effectiveness also has to be considered. Given cost pressures on services, increased expenditure on CSII would mean reductions elsewhere.
Thabit and colleagues suggest that our failure to report details of pump downloads means that we cannot exclude that pump users were failing to make best use of their pumps. In our paper, we explained that REPOSE was designed as a pragmatic trial and that we were unable to explore the detailed reasons why pump therapy was not more successful. We freely acknowledge that patients in our pump arm might not be making the most of their pump therapy but this would be to misunderstand the aims of our trial.
We are somewhat surprised by their suggestion that pump patients require considerably more training and support than patients receiving structured education and multiple injections. As a group with considerable experience in developing and extending the provision of structured education for adults across the UK,  we would argue that the self-management skills of flexible intensive insulin therapy require an equal amount of input and support to calculate basal insulin, estimate the carbohydrate content of food and precise dose of insulin in addition to dealing with hypoglycaemia and exercise etc.
The letter of Thabit and colleagues, plus additional comments on social media have led us to understand that our paper has been used to portray the attitude of the REPOSE group as being against the increased use of insulin pumps, that we are in favour of reducing the funding for this technology and decommissioning pump therapy and we have used the trial data to advocate these views. Nothing could be further from the truth. Scrutiny of the members of the REPOSE Study Group would show that we are enthusiastic about the potential of technology to support people in self-managing their diabetes. However, as we discuss in our paper, the use of technology needs to be targeted and provided to those who can make the best use of it. Merely providing insulin pumps to those who are yet to receive training and be engaged in managing their condition is likely to be ineffective and we believe the results of our trial demonstrate this. As Thabit and colleagues point out, the successful use of insulin pumps requires expertise and a willingness to engage with the details of pump treatment.
It is a sad fact that the vast majority of adults with type 1 diabetes in the UK, are currently, not provided with even basic high quality training in managing their own condition. The implementation of the original NICE guidance TA60 from 2003 (now superseded by the clinical guideline NG17) has been patchy,  and we have reported to NICE that the results of REPOSE reinforce their guidance on structured education. Furthermore, even among those who have attended structured education, only a minority continue to participate actively in ongoing self-management.
We believe that the priority of the diabetes community in the UK should be ensuring that individuals receive more effective structured training and ongoing support, issues we are addressing in ongoing research.  In the meantime, as we conclude in our paper, pumps should be offered widely to individuals who are trained to manage their condition but find that the limitations of sc intermittent injections lead to hypoglycaemia or prevent them achieving tight glucose targets. The ambitious target of 6.5% HbA1c in the latest NICE guideline will increase their number.
On behalf of the REPOSE Study Group
1. Heller S, White D, Lee E, Lawton J, Pollard D, Waugh N, et al. 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. Health Technol Assess 2017;21(20)
2. National Institute for Health and Care Excellence (NICE). Continuous subcutaneous insulin infusion for the treatment of diabetes mellitus. NICE technology appraisal guidance 151 [Internet]. London: NICE. 2008. Available from: https://www.nice.org.uk/guidance/ta151
3. Cummins E, Royle P, Snaith A, Greene A, Robertson L, McIntyre L, et al. Clinical effectiveness and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes: Systematic review and economic evaluation. Health Technol Assessment 2010;14(11).
4. Colquitt JL , Green C, Sidhu MK, Hartwell D, Waugh N. Clinical and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes. Health Technol Assess 2004 Sep 20;8(43).
5. NICE TA 60 Guidance on the use of patient-education models for diabetes (TA60)
6. NICE. Type 1 diabetes in adults: diagnosis and management NG17.
Competing interests: SRH reports grants from Medtronic UK Ltd, during the conduct of the study; personal fees from Sanofi Aventis, Eli Lilly, Takeda, NovoNordisk and Astra Zeneca outside of the submitted work. His institution has received remuneration from Eli Lilly, Boeringher Ingelheim, NovoNordisk Eli Lilly and Takeda outside the submitted work; he is currently Chief Investigator on an NIHR Programme Grant for Applied Research – RP-PG-0514-20013, whose purpose is to develop more effective training programmes to improve self management of type 1 diabetes
We read with interest the REPOSE study which compared two methods of insulin delivery, multiple daily injections (MDI) and insulin pump therapy, following structured education in flexible insulin treatment in sub-optimally controlled type 1 diabetes. The study reports that following similar structured education at baseline, glycaemic outcomes were not significantly different between pump and MDI users, 12 and 24 months later. Treatment satisfaction and some quality of life parameters were better with pump use. We appreciate the efforts made by the REPOSE investigators, however would like to comment on aspects of the study which may be helpful in assessing the clinical and policy implications of this important clinical trial.
The investigators invited 1278 patients to participate. However, prior to this several patient groups were excluded including those with a strong desire for pump treatment; those meeting NICE criteria for pump treatment; those needing a pump in the opinion of the investigator and those with serious diabetic complications. It is unclear what proportion of all patients with type 1 diabetes met the stringent inclusion and exclusion criteria and were invited to participate. These data are important in assessing the generalisability of the findings.
As an insulin pump is an inherently more complex device compared to an insulin pen, more extensive education, user interaction and input are crucial if the benefits of pump therapy is to be achieved. This involves patients undertaking a series of basal and bolus testing to optimise pump basal rates and bolus calculator settings respectively, as well as being trained on advance features such as extended wave functions and temporary basal rates during meals and exercise respectively. Ascertaining patients’ interest in pump therapy should thus be an a priori step, as it currently is in clinical practice. The authors acknowledge this limitation, but by enrolling those who were ambivalent or perhaps even negative about pump therapy, it is impossible to determine if the outcomes are representative of current clinical practice, where patients’ wishes are adhered to.
The per protocol analysis, which compared those successfully completing therapy in both study arms, seems informative because it probably eliminated many unmotivated pump therapy patients. It showed a statistically significant between-group difference in final HbA1c in favour of pump therapy.
Duration of pump training may also explain the small between-group difference in final HbA1c. Based on our own experience and others, to achieve meaningful treatment goals, pump training curriculum would typically take several weeks than one, as was done in the REPOSE study.
The study design aiming for comparable educational exposure, to achieve balance between MDI and pump groups, may have inadvertently caused education and support in pump use to be reduced, compared to that normally provided in usual clinical practice. The premise that pump users should have the same amount of education and patient contact may be misguided, as education on the technical features and developing an ability to act on available information is an integral part of these innovative treatments. This should be adjusted for in cost-effectiveness analyses, but probably not in study designs.
The authors did not provide any device utility data of the aforementioned pump features during the study. Such data can provide meaningful insights of pump usage during the study, as suboptimal pump therapy use may lead to suboptimal clinical outcomes. The unexpectedly high baseline HbA1c (9.3%), given the inclusion criteria in the pump cohort, may indicate a degree of treatment non-compliance or motivational issues within this group, which may also be of clinical relevance. Additionally, it may help readers to assess whether pump use during the trial reflected contemporary clinical practice. Modern pumps and meters are downloadable and allow analysis of information about insulin delivery and glucose control, which can be shared with the clinical care team who then supports therapy adjustments. This element of pump management was not fully explained, and given that HbA1c levels remained elevated and insulin delivery amount remain unchanged, pump management may not have been optimised possibly due to protocol design restrictions to balance contacts with MDI and pump patients.
Participants used two fundamentally different bolus calculators during the study. Those allocated to MDI used the Roche Expert meter, which aims for the mid-point of target range, whilst pump users the Minimed Paradigm Veo bolus calculator aiming for the upper limit of target range. It would be useful to confirm if equivalent glucose targets were maintained, by analysing the devices downloads.
The study was powered to detect a minimum 0.5% HbA1c between-group difference (which was deemed clinically worthwhile) in the sub-group of participants with HbA1c values >7.5% at randomisation. During the study, the point estimate for the treatment effect in those with baseline HbA1c values >7.5% was a non-significant 0.24% HbA1c lowering in favour of pump users. However, the 95% confidence interval for this treatment effect was (−0.53% to 0.05%), which includes 0.5% HbA1c lowering. Therefore, the trial estimate of the treatment effect for pump therapy includes clinically worthwhile HbA1c lowering. We suggest that this important point has clinical and policy implications.
We support the use of HbA1c as the primary outcome measure. However, as sensor-glucose based outcomes were not assessed, the opportunity to evaluate other recognised benefits of pump therapy beyond HbA1c alone such as reduced glucose variability and unrecognised prolonged hypoglycaemia events was missed. There is a growing interest in assessing alternative measures of glucose outcomes, which may have been related to the greater treatment satisfaction reported by pump users in the study.
We acknowledge the great efforts made by the REPOSE investigators to augment the evidence base of this important clinical area. However, there are aspects related to the enrolled population, study design, pump training and pump use, that may limit applicability of study findings on clinical practise and pump reimbursement.
Competing interests: HT declares no competing financial interests exist. MKR reports receiving research funding from Novo Nordisk; educational grants from Merck Sharp & Dohme and Novo Nordisk; consultancy with Roche Diagnostics, Ascensia and Cell Catapult, and owning shares in GlaxoSmithKline. RH reports having received speaker honoraria from Eli Lilly and Novo Nordisk, serving on advisory panel for Eli Lilly and Novo Nordisk, receiving license fees from BBraun and Medtronic; and having served as a consultant to BBraun. LL reports having received speaker honoraria from Minimed Medtronic, Animas, Roche, Sanofi and Novo Nordisk, serving on advisory panel for Minimed Medtronic, Animas, Roche and Novo Nordisk.
We read with great interest regarding the comparison of pump therapy with available multiple injection daily in the management of type I diabetes. It is concluded that both groups showed clinically relevant and long lasting decreases in HbA1c as expected. The important point to be noted is regarding the incidence of diabetic ketoacidosis The number or type of serious adverse of diabetic ketoacidosis was greater in the pump group of 17 vs. 5 in other group. This being a medical emergency is an important determining factor that helps for the acceptability of any regimen.
More patients using pumps had several episodes of 5 were confined to the first year; even this also will became a factor for continuing the same regimen post this severe attacks and may be the reason for non significance in the consecutive years. Three episodes occurred in two participants who switched to pump treatment again proves the fact.
In developing countries where the education and literate level is not par with developed countries pump therapy even if it is clinically fruitful may not be successfully accepted by the people as most of the population fall in middle and low income class where even the available option of implementing the multiple injection is still difficult with increase incidence of DKA and other complications.
Competing interests: No competing interests