These comments are an attempt to summarise the discussions at the manuscript meeting. They are not an exact transcript.

Transcription

1 1-Aug-2016 Dear Prof. Heller Manuscript ID BMJ entitled "A cluster randomised trial comparing insulin pump therapy to multiple injections during flexible intensive insulin therapy for type 1 diabetes - The Relative Effectiveness of Pumps Over MDI and Structured Education (REPOSE) Trial" which you submitted to BMJ, Thank you for sending us your paper, manuscript.we sent it for external peer review again and discussed it at our manuscript committee meeting. We recognise its potential importance and relevance to general medical readers, but I am afraid that we have not yet been able to reach a final decision on it because several important aspects of the work still need clarifying. We hope very much that you will be willing and able to revise your paper as explained below in the report from the manuscript meeting, so that we will be in a better position to understand your study and decide whether the BMJ is the right journal for it. We are looking forward to reading the revised version and, we hope, reaching a decision. Yours sincerely, Tiago Villanueva Assistant Editor tvillanueva@bmj.com **Report from The BMJ s manuscript committee meeting** These comments are an attempt to summarise the discussions at the manuscript meeting. They are not an exact transcript. Members of the committee were: Wim Weber (chair), Tim Cole (Statistician) Jessamy Bagenal, Amy Price, Elizabeth Loder, Jose Merino, Rubin Minhas, Tiago Villanueva Decision: Put points Detailed comments from the meeting: First, please revise your paper to respond to all of the comments by the reviewers. Their reports are available at the end of this letter, below. Please also respond to these additional comments by the committee: - Please address the previous reviewers comments too. - Our statistician made the following comments: A well-designed and executed cluster trial with negative results. My only beef with the design is the choice of primary outcome which excludes data from patients with baseline HbA1c < 7.5%. Effectively they are stratifying on HbA1c with cut-off 7.5% and restricting the analysis to the high group most odd. The abstract does not make clear that the courses constitute the clusters. They should explain the choice of a negative binomial regression model for the number of hypoglycaemic episodes. Table 10 is striking in that the pump is associated with highly significantly greater treatment satisfaction at 12 and 24 months (p < 0.001), yet it is not mentioned in the abstract. So perhaps the abstract conclusions need to be more nuanced. - Several editors were supportive and in favour of restarting the review process (see reviews 1-5). - One editor said the trial conveys an interesting message but he pointed out there are discrepancies in outcomes between register, protocol (the 2015 version ), and the paper, that you would have to address. He added that patient selection seemed a bit puzzling: Why exclude patients meeting NICE criteria for pump therapy? And why not exclude patients with HbA1C < 7.5%, as the primary outcomes does not concern them? - One editor felt the paper conveyed important information but she agreed with the limitations highlighted by a previous reviewer about the need to be careful about generalizing to all pumps and the use of a particular type of glucose meter. In your response please provide, point by point, your replies to the comments made by the reviewers and the editors, explaining how you have dealt with them in the paper. ** Comments from the external peer reviewers**

2 Reviewer: 1 The paper deals with a very important topic on type 1 diabetes management: the comparison between continuous subcutaneous insulin injection therapy and multiple daily insulin therapy. More precisely, it compares the effectiveness of insulin pumps to multiple daily injections for adult patients with type 1 diabetes in terms of glycaemic control, hypoglycaemia risk and quality of life, with both groups receiving equivalent training in flexible insulin therapy. In fact, author s hypothesis was that much of the benefits of insulin pump therapy is not determined by the mere technicality of the pump therapy, but it is due to the improvement in the ability to manage the disease in patient s everyday life. Authors conclude that the structured training and education to diabetes management is the most important tool that lead to a long lasting improvement in glycaemic control, rates of severe hypoglycaemia and psychological measures in type 1 diabetic individuals and that, adding insulin pump therapy to structured training does not significantly enhance educational benefits. We think that the trial design (pragmatic, multicentre, open label, cluster randomised controlled trial) is novel and original for this topic. Moreover, the statistical methods used for the study appear to be elegant and appropriate as well. Finally, clinical relevance of author s findings is strong, being the number of randomized patients very high (especially when compared to previous published studies on the topic), the duration of follow-up long (2 years) and the two studied groups very well characterized. That said, some points needs to be addressed by the authors: Major revision 1) Authors should mention (in the text and/or within one of the tables) how many patients of the insulin pump group did use continuous glucose monitoring systems. In fact, even if the study design takes into account only the insulin delivery systems (MDI vs. CSII), we actually know that modern insulin pumps (like the Medtronic Paradigm VEO model used for the trial) are equipped with sensors measuring glucose concentration from the interstitial fluid and that this combined approach (sensor augmented insulin pump therapy - SAP) is till today considered the gold standard for type 1 diabetes treatment. Authors should, therefore, mention how many patients did use sensors (either with the retrospective or with the Real-Time approach) in the insulin pump group and eventually the percentage of time of sensor use. We believe that this point is crucial because it could directly affect the primary endpoint of the study (change in HbA1c after 2 years). 2) The definition of moderate hypoglycaemia as an episode which could be treated by the individual but where hypoglycaemia caused significant interruption of current activity leading to impaired performance or embarrassment or being woken during nocturnal sleep and of severe hypoglycaemia as an episode leading o cognitive impairment sufficient to cause either coma or requiring the assistance of another person to recover, although detailed, should be supported also by a cut-off value included in the definition: for example plasma glucose <70 mg/dl for mild/moderate hypoglycaemia and <50 mg/dl for severe episodes. In fact, most of the randomized trials assessing the rates and severity of hypoglycaemic episodes in type 1 diabetic patients do consider at least one cut-off value to define this adverse event (hypoglycaemia needs to be proved by definition by a low plasma glucose value measured at the time of the symptoms). 3) The time-log of the trial included four main visits (baseline and after months) in which primary and secondary endpoints have been measured. We ask the authors if between the scheduled visit at 12 months and that one at 24 months there have been other intermediate visits or contacts with the patients in which caregivers could have changed insulin dosages or reinforce education on diabetes management. If not, we firmly believe that 1-year gap between two consecutive visits is too much time for any type 1 diabetic patient. Could it be the explanation for the modest improvement of the patients (in terms of HbA1c reduction) between the one-year and the two-years follow-up visit (Figure 2)? 4) The participants of the study were adults with T1DM willing to undertake intensive insulin therapy, with no preference for pumps or MDI. Did the authors consider a minimum number of plasma glucose measures per day to be the patient included in the trial? Did the two studied groups have similar baseline number of plasma glucose measures per day? Please specify this data in the methods section. Patients with poorer compliance to self-monitoring of blood glucose (even if well educated to the management of their insulin therapy) could, in fact, be less likely to get an advantage from insulin pumps. 5) In Table 1 (baseline demographics), baseline characteristics of the participants have been reported. Beside the rates of macrovascular complication, also the rates of microvascular complication should be reported in that table. How the high rate of macrovascular complications (55.1%) in this relatively young cohort of patients (mean age 40.7 years) could be explained? Could that population be considered at higher risk for cardio-vascular diseases? It is an important point because patients having a higher cardiovascular risk profile (like those with previous myocardial infarction) should have a different target range of HbA1c. 6) In the conclusion section of the abstract (page 2), the authors say, adding insulin pump therapy to structured training in flexible insulin therapy did not significantly enhance educational benefits on glycaemic control, hypoglycaemia or psychological outcomes in adults with T1DM. We think that the authors should mitigate this conclusion given that Tables 7-10 show an improvement of several parameters of quality of life with CSII vs. MDI (for example the significantly improved treatment satisfaction of patients on CSII vs. MDI detected by DTSQ, together with less restrictions in diet and daily hassles in the DSQOL scale). We believe that these statistically significant data should be interpreted as a meaningful result achieved by the insulin pump therapy group (not only as a modest quality of life benefit ).

3 Minor revision 1) In the abstract section (on page 2, line 5) the abbreviation ITT, that appears for the first time, should be defined (like the authors did for MDI and T1DM on page 1). 2) We believe that the effectiveness of the intervention tables, assessing the proportion of participants with HbA1c 7.5% at 6-12 month (Table 4) and at 24 months (Table 3), could be joined together. 3) Do the authors have data on the changes in glycaemic variability between the two studied groups at the end of the trial (e.g. LBGI, HBGI and/or MAGE indices)? 4) Table 7, 8, 9 describe quantitative psychosocial outcomes in terms of mean difference in change from baseline at 6-12 and 24 months. Many data and numbers (either significant or not significant) are presented in those tables and the reader may find them repetitive and hard to interpret. We advice the authors to summarize some of this results and/or underline statistically significant differences (for example using bold characters). Additional Questions: Please enter your name: Andrea Tumminia Job Title: MD Institution: Department of Clinical and Experimental Medicine, University of Catania (Italy) A fee for speaking?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No in any way (please see BMJ policy) </a>please declare them here: Reviewer: 2 GENERAL COMMENTS In this work, the authors compare two groups of patients with type 1 diabetes, managed respectively with insulin pumps and multiple daily injections (MDI), who received equal training in flexible insulin therapy. Moreover, they analyze a number of psychological measures, to gain further insight information on treatment satisfaction and quality of life. The results of the psychological measures analyzed may be of interest to both patients and clinicians, and their discussion should be improved. METHODS/CONCLUSIONS Out of a total number of 1278 invited patients, respondents were 885 (69.2%), but 523 (59%) denied being interested in participating. As a result, participants were 362, i.e. 28.3% of those invited, accounting for those who were interested. Since in the Discussion the authors state A common reason given by patients for not wanting to participate in REPOSE was reluctance to use an insulin pump, this could determine a selection bias. The authors also write: Our hypothesis was that much of the benefit of pump therapy may come from the re-training and education in insulin use given to enable patients to use pumps safely. Since the trial assessed treatment efficacy both in patients using pump therapy and in those using MDI, who received equal training in the two methods of insulin administration, both sets of participants should be taken into consideration in the conclusions. Moreover, can the authors explain in the Discussion, why the trial was not suspended after a year, given that interim analysis found that the intervention was ineffective (tables 2-5, figure 1)? The authors also report that most episodes of ketoacidosis were due to infections. Which infections? RESULTS/DISCUSSION The Quantitative psychosocial outcomes are perhaps the most interesting findings of the study, but they are given little prominence in the Discussion. These data are highly relevant for patients as well as clinical practice and should be discussed.

4 Additional Questions: Please enter your name: Emma Job Title: Epidemiology Unit Institution: University of L'Aquila A fee for speaking?: No Funds for research?: Yes Funds for a member of staff?: No Fees for consulting?: No in any way (please see BMJ policy) </a>please declare them here: I declare no competing interests Reviewer: 3 The conclusion that high quality, structured self-management is critical to the management of diabetes substantiates many previous studies. It is not a surprise that with a good training program, equal patient contact, and the use of bolus calculators in both groups that there would not be a difference in A1c levels between the two groups. This is an important paper, and a large randomized trial that needs to be published and is of the quality to be published in the BMJ. There are some details about how the study was conducted that should be clarified in the report, particularly related to the training and management of subjects assigned to pump therapy. This study was conducted at very experienced diabetes centers where the standard of care is to use MDI therapy. The DAFNE program is an excellent, well established diabetes education program for teaching the principals of diabetes management and insulin dosing for meals based on carbohydrate counting. Prior to this study it is unclear whether the standard of care involved using a glucose meter which included a bolus calculator. Since this is a major study which will have significant impact on the ability of patients to obtain coverage for pump therapy, it is important to understand how well pump therapy was also managed during the study. Since pump therapy is used in a small minority of patients in the UK, it is critical that the patients and health care providers were providing state of the art pump education and management. There is no description of the specific information and training provided to the pump group in this paper, or the paper referenced on the study design, or in the study protocol that was attached. It is difficult to evaluate the differences in therapy between the two groups without a description of the education and selfmanagement skills provided to the pump group. This is particularly important since the MDI group was not a typical MDI group since they used a meter with a built in bolus calculator which included CHO:I ratios, correction factors and active insulin calculations. These are many of the features that often distinguish pump use from MDI use (particularly the active insulin calculation). It is therefore important to understand what was taught to the pump group which distinguished the pump therapy from the MDI therapy. Specific questions are: 1) How were basal rates determined - were there systematic 3 AM glucose checks to help determine overnight basal insulin delivery; 2) How were basal rates determined during the daytime since there may be varying insulin sensitivity throughout the day with diurnal patterns; i.e were meals withheld, or were non-carbohydrate meals consumed to determine daytime basal rates in the morning, afternoon and evening 3) What was the recommendation for ketone testing to assess for infusion site failures; 4) What was taught on how to detect an infusion set failure: i.e. testing of serum ketones, failure to correct by 50 mg/dl in 1 hour with unexplained hyperglycemia? 5) Were subjects given blood ketone meters; 6) Were subjects taught how to download their meters and pumps into CareLink; 7) Were subjects instructed on how to deliver dual wave boluses for meals with high amounts of protein and fat, or meals traditionally associated with delayed absorption (such as pizza, burritos, etc.). In a dual wave bolus, how much was given as a normal bolus and how much as an extended bolus and what was the typical duration of an extended bolus. Was this determined based on the protein and fat in the meal?

5 8) I assume CHO:I ratios were determined based on the DAFNE program. Was there systematic testing of results following a meal with a known carbohydrate content? (In the results they could provide the differences in CHO:I ratios at breakfast, lunch and dinner between the MDI and pump groups.) 9) What was the teaching for exercise to deal with prolonged suspension of insulin delivery and what was the recommended insulin dosing for exercise carbohydrates ; and 10) What information was giving about issues with wearing an insulin pump and intimacy? The other major question for those receiving pump therapy was how was each pump visit conducted? Was the CareLink report downloaded and reviewed with the subject so both the subject and physician reviewed the reports looking at: summary statistics, the modal day, pump settings, and daily details to determine how well correction doses and CHO:I ratios were working and if adjustments needed to be made to basal rates and the approach to exercise. Did each center download and review reports before each visit. Was there a difference in which Carelink reports were used at different centers, or was this standardized? 1) It would be helpful to know some of the basics of the insulin pump therapy provided in this study: 1) How many basal rates did patients typically use each day?; 2) what was the basal to bolus insulin ratio; 3) How frequently were infusion sites changed (data from the pump download, not by patient report); 4) what was the usual active insulin time and how did this differ when compared to the MDI group; 5) what were the typical number of pump suspensions each week?; 6) what were the mean number of boluses each day for the pump group compared to the MDI group, and was there a difference in the number of correction boluses between the two groups; 7) Were pump and meter recommended insulin doses overridden? 8) What were the insulin sensitivity factors and blood glucose targets throughout the day and night. Did they differ between the MDI group and pump group? All of the data mentioned above is available in a typical CareLink download. In the US the TID exchange reports that pump use in adults is now over 60% whereas in the UK pump use is at 6%. Were there any centers that had a high level of experience with pump use, i.e. with over 50% of their patients on pumps? Was there a center effect between centers who had high pump usage compared to those with low pump usage? The higher rate of DKA in the first year of pump therapy which resolved by the second year of pump therapy may indicate that subjects could have benefited with more extensive training on detection of early infusion set failure at the onset of the study. This may relate to health care providers being relatively inexperienced with insulin pump therapy which may have potentially biased the study. This potential bias is not addressed in the paper. In the US there are many support groups for pump users both virtual and through social media. Was this social networking also available to the pump users in this U.K. study? Is this something that naturally developed as a result of the DAFNE training groups? To compare the two study arms one of the most important factors is the frequency of blood glucose monitoring in each study arm. This could be provided in a table which would include some glycemic outcomes, such as mean blood glucose, % of readings <70 mg/dl. Were any subjects using a continuous glucose sensor? One area where pumps may be of benefit is in the reduction of nocturnal hypoglycemia. Currently the best way to assess nocturnal glycemic control for insulin dose adjustments, and for the incidence of nocturnal hypoglycemia is with continuous glucose monitoring. For a randomized clinical trial a periodic week or two of blinded CGM wear can provide important statistics on mean blood glucose levels, and % of glucose values in range, and in the hypoglycemic range, and daytime and nighttime control can be separated. An A1c does not capture hypoglycemic risk, and the number of events may be significantly underestimated in those with hypoglycemic unawareness, and in those with nocturnal hypoglycemia. Was any consideration given to using a blinded CGM in this study, or in a subgroup of the subjects? Additional Questions: Please enter your name: Bruce Buckingham Job Title: Professor of Pediatric Endocrinology Institution: Stanford A fee for speaking?: No Funds for research?: Yes Funds for a member of staff?:

6 Fees for consulting?: Yes in any way (please see BMJ policy) </a>please declare them here: I have received research support and I am on the medical advisory board for Medtronic Diabetes. I have received research support and have been on the advisory board for Tandem. I have been on an advisory board and will be receiving research support from Insulet. I have been on an advisory board and will be receiving research support from Animas. All of the research support and advisory board functions related to closed-loop therapies. Reviewer: 4 The REPOSE study group investigated the impact of insulin pump therapy vs. multiple injections in patients with type 1 diabetes. Using a pragmatic, multicentre, open label, rct setting both treatment arms were provided with similar training efforts to avoid inequalities of attention and to detect the actual effect of pump therapy. As compared to baseline HbA1c levels a reduction of 0.5% was observed after a most relevant follow up period of two years with no significant difference (0.2%) between the two treatment arms. Of note, only every forth patient had adequate gylcaemic control after 2 years as defined by a HbA1c level <=7,5%. In addition, a reduction of severe hypoglycaemic episodes with no differences between treatment arms was observed. A quite extensive evaluation of psychological outcome measures showed improvements for some domains in favour of pump users. The results are of significant relevance and highlight the fact that the implementation of pump therapy only is unlikely to overcome the barriers of inadequate diabetes control in the majority of suboptimal controlled patients. In the manuscript presentation of data and interpretation of the results are well-balanced. I have some comments to the authors: - Pre-existing treatment regime may have an impact on the outcome measures. The introduction of intensified insulin therapy could lower HbA1c levels and hypoglycaemic rates. Thus, how many patients in the two groups were practising intensified insulin therapy, twice daily regimes,... before entering the trial? Did pre-existing treatments have an impact on outcome variables? - What was the definition of optimised MDI treatment (p 5, 23), the research protocol does not clarify this exclusion criterion. - Insulin levels decreased slightly in the pump group, whereas the dose remained relatively unchanged in the multiple injection group. Nevertheless, HbA1c reductions were observed in both groups underlining the effect of appropriate insulin administration after structured education. However, shortage of insulin dose could be one reason for the documented ongoing suboptimal control in a majority of patients (and the incidence of hypoglycaemic episodes). Please provide baseline insulin dose to allow better interpretation of the results. - On page 6 the authors emphasise the importance of achieving an adequate HbA1c level without experiencing severe hypoglycaemic episodes. What was proportion of patients with HbA1c levels < 7,5 AND no recent severe hypoglycaemic episode (12 month) at baseline and, in particular, during the interventional period for the both treatment groups, respectively? - Figure 4 does not include severe hypos. Did patients with anamnesis of severe hypoglycaemic episodes better (HbbA1c, severe hypos), if randomised to pump therapy? - The authors discuss the change of the NICE HbA1c recommendation from 7.5% to 6.5%. In my opinion it would be worthwhile to present the number of patients, who actually achieved this elusive goal (HbA1 <6.5 with or without severe hypoglycaemia) in the current investigation. Johannes Plank Medical University Graz, Austria Additional Questions: Please enter your name: Johannes Plank Job Title: Consultant Internal Medicine and Endocrinology Institution: Medical University Graz A fee for speaking?: No Funds for research?: No Funds for a member of staff?: No

7 Fees for consulting?: No in any way (please see BMJ policy) </a>please declare them here: Reviewer: 5 The REPOSE trial addresses an important research question, namely the potential benefit of insulin pump (CSII) therapy compared to multiple daily injections (MDI) on top of structured training (DAFNE) in patients with type 1 diabetes with no strong preference of treatment mode. The dilemma in most previous trials in this field is as pointed out by the authors that the CSII patients often receive much more attention in order to take out the full benefit of this complex therapeutic option. The question in the case of benefit of CSII is then whether it is explained by the treatment itself or by the training. The REPOSE trial tried to circumvent this problem by applying a similar amount of training in both treatment arms. However, then the question is: Did the CSII patients receive sufficient training and attention throughout the study to harvest the full CSII effect? This deserves further attention in the limitations section of the discussion (see also pt. 2+3 below). Overall, the trial is well-powered to assess its primary endpoint and the two-year duration is a further strength as it ensures capture of potential exhaustion after an initial enthusiasm following a new intervention in subjects only moderately motivated (applies particularly to the CSII group). Given the resource consuming training program and subsequent randomized insulin intervention, the overall primary outcome in the trial is rather disappointing and far from the therapeutic target for T1DM. The reason for this is probably that no insulin titration was performed, emphasizing that educational efforts cannot stand alone in the strive for better glycaemic control. As there are considerable numbers of patients with T1DM that do not reach glycaemic target despite structured education, I find the study very important for clinical practice as the choice of the next intervention based on these data obviously should be individualized. The manuscript is well-written, but the conclusion needs reformulation. The authors attribute the general study result to the DAFNE training which cannot be done as there is no control group for this intervention (could as well be a general trial effect). A more appropriate conclusion in alignment with the title and the aims and supported by the data would be: There was no difference in the effect of analogue-based MDI or CSII applied without titration when added to structured training in patients without preference for any of these treatment modes. In order to fully understand the results a number of questions need to be addressed and clarified in the manuscript: 1) What was the baseline treatment of the cohort? Specifically, how many were already on twice daily detemir? 2) How often were the patients seen during follow-up and were ad hoc contacts with the purpose of optimizing (particularly CSII) therapy allowed? In other words: did the CSII patients obtain the full effect of the treatment? 3) In the same line: are there any data from device downloads to assess the actual compliance in the two groups regarding carb counting and bolus calculation? 4) To understand the HbA1c outcome it is important to know the insulin dose at follow-up broken down to bolus and basal and a description of the principles for insulin adjustments during the trial. 5) In addition, some data on glucose measurements would be of value, F-PG in particular. 6) Why did the trial include subjects (with HbA1c < 7.5%) that could not be assessed according to the primary endpoint? In clinical practice such patients without strong desire of CSII treatment would only be considered for CSII in the case of problematic hypoglycaemia was this the case here? From a glycaemic perspective these subjects could contribute to differences in hypoglycaemia but they are rather few and there are no power considerations in terms of hypoglycaemic endpoints. 7) Regarding the hypoglycaemic endpoints, severe hypoglycaemia was defined according to the ADA guidelines, requiring no confirmatory glucose measurement. The second milder hypoglycaemic measure, moderate hypoglycaemia likewise required no biochemical confirmation. This is in contrast to the ADA recommendations that requires confirmation of all milder events in trials. Finally, it would have been reassuring to have an estimate of asymptomatic hypoglycaemia based on self-monitoring or blinded continuous glucose monitoring.

8 8) The authors state that participants had no preference for CSII or MDI. How was that assessed? And how large a fraction of the population was eligible according to this criterion? It might be expected that most patients have some degree of preference. 9) What was the rationale for only allowing detemir twice daily as basal insulin in the MDI arm? In the discussion is mentioned a need for studies comparing CSII with best MDI. Do the authors consider detemir twice daily as the best basal insulin therapy? Ulrik Pedersen-Bjergaard Additional Questions: Please enter your name: Ulrik Pedersen-Bjergaard Job Title: Consultant, Ass. Professor Institution: Nordsjællands Hospital Hillerød and University of Copenhagen A fee for speaking?: Yes Funds for research?: Yes Funds for a member of staff?: No Fees for consulting?: Yes in any way (please see BMJ policy) </a>please declare them here: Speaker fee: AstraZeneca, Novo Nordisk Advisory panel: MSD, Novo Nordisk, Sanofi Aventis Research support: Novo Nordisk